AI agent that recommend board games based on user preference

Game Night are exciting, until it time to pick a game. The hardest part of a Game Night isn't finding good game. It is agreeing on what to play.

Comment

Inspiration

Jason has been into board games since he was a kid, but really got into modern board games around 2017 starting with Betrayal at House on the Hill. Since then, he has built up a collection of almost 200 games and plays pretty regularly with a group.

There’s always been one problem though. Picking what game to play.

Even when everyone is excited to play, the group ends up spending way too much time just trying to decide. People want different things, some want something quick, others want something heavier, and it turns into a back and forth that eats into the night. And that doesn’t even include setup time.

Jason has also seen the same issue come up a lot in board game groups online, so it is clearly not just one group.

He brought the idea up to his teammate Ky, and Ky immediately understood the problem. He has also played a lot of board games with his siblings and had the same experience. We know there are apps for tracking games and learning rules, but nothing that actually helps a group decide what to play based on their own collection.

So we decided this is a perfect thing to try and solve.

What it does

An AI agent that takes user input and determines the best possible board games based on number of players, complexity, theme, and playtime. The agent remembers context throughout the conversation so users can refine their request naturally without starting over.

How we built it

Built on the n8n platform using the Google Gemini Flash AI model. We implemented a scoring system inside the AI prompt to rank games out of 8 points based on player count, complexity, theme, and playtime. A memory node was added to retain context across the conversation. The AI parses natural language input and explains why each recommended game fits the group.

Challenges we ran into

-AI model was using too much request and token -Slow response time -The scoring system was not working as intended

Accomplishments that we're proud of

-Figure out a way to implement the scoring system better -Shrink down the complexity of the workflow -Reduce the time it take for workflow to run

What we learned

-The AI model doesn't need to be overcomplicated -Have a focus on efficiency

What's next for AI agent that recommend board games based on user preference

-Connect BGG API to the AI agent so that it can directly access the board games data -Implement a profile feature, where the user preferences can be saved, and when the users ready to play they only need to input the amount of players, and the amount of times they want to play

Built With

Share this project:

Updates