Documentation

Discussing around LLM, I decided to play around the specific words with these models so I tried to make a word riddle engine with Gemini API.

So I gave this prompt below.

But I found the last sentence that ask it only to return pure JSON not working sometimes so I did a bit change in dealing with the accepted data.

The next step is just some button and logics that I asked GPT for help.

And the functions for all these buttons.

Below is a small video demo!

录屏2025-11-07 上午11.20.51.mov

The founding is that the clues that Gemini generated is somehow pretending to be mysterious, sometimes this is very confusing.

Some improvements for the future is:

1. UI UX experience: I didn’t work too much on the UI/UX this week so definitely need to be improved.
2. Score board: make it more like a game by setting certain amount of times and keep track of the words user can get.