Lol. After 5-6 questions it started doing the opposite response that I choose. I choose A every time and it started forcing me to take B option. Classic LLM success.
It’s doing so because it includes the entire history of choices in every prompt. By using the API directly and modifying the response to only include the assistant messages and not the user messages it would stop this behavior.
You can increase the number of options too, I increased them to 4, made it more interesting (You can literally just type: Let's continue but with 4 options)
Stop including the Fantasy Online line in the output.
Stop including link to "Join our discord..." in the output
Makes ît marginally faster to respond