Well, thank you for the explanations.
Game has it’s support for external API, because it was the first that I did, but it looked more censored than this one I got.
I’ve read about Jailbreak, the idea is fine, but it needs some actions from the user. My current LM is 3B and it has a lot of rules inside my game, so it may look stereotypically answering, it may be due to the reputation or something, when it is pure without prompting or with encouraging prompting as will happen after reputation > 25, it will look better. Also if I will use 7B it will be talking much better if I will see that everyone get boring answers, I will do it, but most probably I will then switch to GPU usage and add API capabilities and disabling local LM if API key is used.
And about LM censorship. The small models do not have a lot of data in fact, so they are limited by nature, they are able to maintain dialogue and to react more or less naturally. But they usually do not have fetish literature learning or something, what LLM have (Anthropic, Perplexity, OpenAI). I’ve tried SLMs with the fetish Loras, they are acting so unnatural, like they start fetish talks at any point without even following context.
Thanks for the idea about API, because I thought it is too complicated for users, if not, I can do it really fast even with the current version just by adding API input at start (choose API vs local SLM)
I bought the game, but I’m really struggling to every have it take actions like cooking. Instead, the AI hallucinates and chats that it’s cooking, but it won’t do it in-game. The character will sometimes eat from the inventory, but I keep seeing this message
Also, after sleeping, the fullness gauge breaks and stops decreasing, which more or less ends what you can do and requires making a new save to fix. The other gauges look like they could have issues as well.
It’s a promising idea, if only the keywords were a little more clear and if the gauges didn’t break.
Thank you for the feedback! I’m fixing all the reports I got, hope I will release update soon.
Fixed some bugs and added API capabilities, it is really cool. I will make the built in LM to use GPU also, so fast. Still testing, have not released yet
Hello guys!
I’ve made a separate version for the GPU/ API LM support. Now chat works fast and also API should work with the most common providers and with custom setup, LM Studio for example or anything in fact. I’ve tested both local with GPU and API with LM Studio. So you may try it and tell me if it is any better 
Vulkan backend for LM means it will work with any GPU, like Nvidia, Radeon or Intel. I think best compatibility is important. Tell me how it works on different GPUs, please.
So again, in this version LM works with GPU, CPU is not used, RAM is not used, GPU is used.
Who bought already may try new version, tell me what changed from your perspective.
Got a chance to try the update. The GPU backend makes it run much faster and the character feels more responsive and seems to have a better idea of context, nice job.
However, the fullness and other gauges still break almost immediately, so it’s more or less unplayable.
Thank you for the feedback. What do you mean by breaking gauges?
Her fullness stops going down is the main issue. This prevents any food-related activities because her fullness value will never decrease.
The UI for health, hydration, and others will also go blank, but it’s unclear if these are the actual values or an error since it happens suddenly.
Thank you, I will fix these things
Should be fixed in 0.2f1, should be active in a few minutes
Still has the fullness meter break, especially after sleeping. My computer is running an Nvidia 5070 and has 32GB of ram, so I think it should be capable of running this game. I think it’s an issue with the coroutines you’re running for handling decay.
Could you explain how it breaks? It is supposed just to go down during sleep, what happens in your game?
Here’s a 36s recording, sped up a bit to show the fullness gauge not moving, along with abnormal display on a few others:
It was working fine during sleep, but after waking up and eating, it no longer decreased. Waiting for the next sleep cycle has no effect, it seems to stay broken once the fullness decay stops.
Ok, I see, thank you! I will check everything and try to recreate the bug.
Quick question in the near future will there be game of yours full releases for free that getting me intrigued
Is this game compatible with Victus I’ll wait until the full release comes out then purchase it
And my third question will this be able to play on cross platform on Android and iOS
I’m not planning to release free games, sorry.
It depends on the Victus configuration, RTX 4070 (desktop) and 32 GB RAM seems working fine if you get it like this, but maybe with 4080 mobile at least, you may not experience too many problems. But be sure, laptops are getting hot running both 3D game and language model same time
For this game it is technically impossible today. Mobile platforms do not have enough power for it.