???
I… wasn’t upset. I’m still not.
As an aside, I only just realized your avatar is a nice landscape image of a river going through hills. From the tiny image on my phone I thought it looked like some cartoon mouse for some reason… I really need new glasses.
You seem to be inferring a lot of things that I didn’t say.
Talking about how you’re not in the drama and avoiding the drama is not, in fact, avoiding the drama. My point was that your post (perhaps unintentionally) came off as extremely sarcastic. If you are truly staying out of the drama, then why did you post what you did? You could have simply said nothing and let this person show their appreciation.
Weren’t you the one that made the post that this is in response to?
Stable Diffusion is an image generator engine, not a text completion engine.
The text completion engine (AI Text Plugin) is likely a fine-tune of Llama 2.5.
That’s a nice screenshot. Did you forget to include a link so we can check it out?
How is “It’s AI” a tip?
I have been having this same error for over a month. I figured it might be isolated and didn’t give it much thought since I keep all my work in a git repo.
That is really awesome news!
Do you have any plans to open up some of the AI settings, such as temperature and token count? It would be really helpful when setting up different kinds of text generation.
Thank you.
This page on Custom Code does a great job giving all the important info about the Json data of a character.
Your idea to write to the reminder of a character seems viable. Perhaps just a collection of keywords in the reminder? Then, each character bio could contain a description + context on what those keywords mean?
Another option could be something similar to the Strict Game Master that uses code to track things like inventory and skills.
Yet another could be leveraging /mem. That is in chronological order (I think) so maybe you can add entries there?
I would assume it will be as soon as they are able. Last I heard, they were having some issues getting costs down and proper optimizations in place for the text-to-image model. They also said that it could be ready as soon as April, assuming everything went smoothly with the text-to-image upgrade.
Essentially, sit tight. It will be done when it’s done.