/aicg/ - A general dedicated to the discussion and development of AI chatbots#162 - koishi edition▶https://rentry.org/aicg_vg_memo▶This general is for discussion of the interactive experiences you can build on top of AI chatbots.▶This is not a tech support thread. If you need help with accessing the API or setting up one of the frontends, please visit the appropriate thread on >>>/g/.>NewsOpenAI has released the latest gpt-4o-2024-08-06 snapshot which supports Structured Outputs and chatgpt-4o-latest which will continuously be updated to the ChatGPT version to the APIAnthropic will be adding prompt caching soon https://www.anthropic.com/news/prompt-caching#/aids/Llama 3.1 8B/70B/405B and Mistral Large 2 released https://ai.meta.com/blog/meta-llama-3-1 | https://mistral.ai/news/mistral-large-2407additional info: https://rentry.org/aicg_extra_information>FrontendsSillyTavern: https://docs.sillytavern.appRisuAI: https://risuai.netAgnai: https://agnai.chat | https://rentry.org/agnai_guides>Botshttps://characterhub.org | https://chub.aihttps://realm.risuai.nethttps://char-archive.evulid.cchttps://rentry.org/meta_bot_list>Modelsjailbreaks: https://rentry.org/jb-listinggpt: https://platform.openai.com/docsclaude: https://docs.anthropic.com• https://rentry.org/how2claudegemini: https://ai.google.dev/docslocal: >>>/g/lmg• https://rentry.org/meta_golocal_list>Botmakinghttps://rentry.org/meta_botmaking_listhttps://desune.moe/aicharedhttps://agnai.chat/editor>MetaOP templates: https://rentry.org/aicgOPaicg botmaking events: https://rentry.org/meta_event_listlore: https://rentry.org/aicg_chroniclescard v3 status: https://rentry.org/zdwfrgqqservices assessment: https://rentry.org/aicg_metalogs: https://sprites.neocities.org/logs/reader | https://chatlogs.neocities.orglatest proxies: https://rentry.org/desuproxyrebornPrevious: >>493751493
>ANCHOR
The people's voices were not heard. This should have been an A-Team SPH thread that drains /aids/ users.
>>493939842She has an egg on top of her head.
any koishibots?
>>493939862#1 trending bot of chub right nowhttps://characterhub.org/characters/Babanon/lisa-your-clumsy-mom-5e73f552e55bBasically the same as one of the most downloaded bots, but just slightly different
>>493943085too much coom right off the bat. Maisha is better.
>>493941521guess not
>the state of chub
>>493952015>judging a book by its cover/lack of cover
>>493952418yeah, let me just read the gay dino prison book
>>493952015*blocks*
My text gen isn't continuing on from the last sentence properly, what am I missing? Using Nemomix Unleashed in LMstudio, all default.
>>493952854It's working perfectly fine for me. A card issue maybe?
>>493953009All the other models I've used work fine but something's up with Nemomix, I don't want to have to give up on it though, since it's the best local model I've tried.
>>493952015I can't believe there is an actual need for a filter by "has avatar". The absolute state.>>493952418Kek
>>493952854I assume you're talking about the continue feature on ST? This probably means there is some extra prompt being added at the end, like a jb/post history instructions prompt.Look at the console and see what it is sending to the model.
>>493953459Are you using the mistral context template that came with the model?
In case you haven't realized, you are using a Fillyfucker thread here.He has puts aids in the OP and he is impersonating Kalakan. Move to the other thread >>493942804
>>493954174My brother YOU are the fillyfucker. Stop trying to incite unnecessary drama.
>>493954174This thread has more replies so this is the real thread
>>493953834No? I can't even find what you're talking about.
>>493954174I did ask nicely, though. >>493939468
>>493955214https://huggingface.co/MarinaraSpaghetti/SillyTavern-Settings/tree/main
>>493955408People are sharing ST settings via huggingface? Man the local scene is something else.
>>493955023The other thread has yuri logs so it's the real thread.
>>493955739No one is really doing that, but having git and the rendered readme is neat, I guess.
>>493956545This is a sloppy yuri thread now.
Sonnet 3.5 repeats itself too much, how do I stop it?
>>493956545
>>493958261just tell it not to repeat itself
>>493958261Try the usual stuff, I guess.https://rentry.org/how2claude#fighting-the-repetitionThe big thing is that once it sets in it's encouraged to keep doing it (because LLMs fucking love repeating patterns), so you need to shake things up and force it to do other stuff for a while. Which, yes, can include just outright telling it "hey, stop talking about her pensive expression in every second paragraph".
>>493958261Spot the repetition early on and swipe it away as soon as you see it. Cobble multiple replies together if you must.
>>493962921I do this. I still find Sorbet has a certain way to start responses which is worse than {{char}} spam.>[Verb] [adjective], {{char}}...
>>493963570Another thing that helped me is to switch to 2.1 for a couple of replies. It breaks up the monotony.
>>493963834I find myself switching on every other response then. So the ultimate solution is to opt into sorbet only when it's intelligence is needed
>>493964035Pretty much, yeah. Sorbet sucks at prose, no matter what you try to do.
