SuihtilCod
commented
6 months ago After further investigation, I've found that the reason for this development is not because of the way I'm typing, but rather because the bot's "Chat Direction" memory has been wiped. Why, I don't know. Regardless, I'm closing this specific issue because it's no longer relevant or accurate. My apologies.
Superseded by Issue
-
270
User information
Login Type: Apple ID
UserID: ZWJPhmAYFGNAaeAVpqAm4iF9r1j2
Actual behavior: Unless the user uses a Chai-style response, the bot almost always switches roles and acts as though the user is the bot and vice-versa, or on occasion, breaks character entirely.
This one has me baffled and frustrated. I recently created a bot based on a specific character from a cartoon show. After some testing, I've found that if I use a Chai-style (or "IRC-style", I guess?) of responding – where asterisks mark actions and plain text is spoken – the bot will function as normal. However, if I try "book-style" – asterisks for emphasis, spoken words in quote marks – then it stops responding normally.
I have no idea why it's doing this – especially after a very delightful training conversation. Very frustrating after all the effort I put into it. No other bot that I've interacted with, whether I made them or not, has ever had this problem.
Provide Screenshots if possible
Example conversation with the bot in question.
The bot swapping identities after a book-style post.
The bot completely breaking character and swapping identities after a book-style post.
The bot successfully staying in character after getting a Chai-style reply.