When you travel, npcs you've recently interacted with may follow you at the discretion of the narrator. This is much more likely to happen if you describe the NPC as travelling with you to the new location.
MoralHazard
Creator of
Recent community posts
I'm not able to reproduce this error, unfortunately. If you're willing to try again and check the dev console I can probably fix it, but as of now I'm not sure what would cause this; perhaps something specific to your local environment.
There is a chance this was caused by errors that could previously occur when buttons were pressed while another function was executing. If so, the problem should be resolved now.
I appreciate that I'm supposed to wait a long period of time before posting about a deindexing, but reading this forum it seems like the squeaky wheel gets the grease, so here's me squeaking.
All my games were delisted a few days ago, as soon as I added generative AI tags. I've put in a support request--no response yet. I can't help but think this is a very punitive approach to managing AI generated content; by voluntarily and proactively complying with itch.io's tagging request, I've been automatically delisted for an indefinite period of time. My games are not mass produced "AI slop", the simply use LLM access to add dynamism to the games.
Thanks! I've (probably) figured out the issue. Did you submit text without adjusting any values in the in-game app? I intended for users to adjust the app before their first text submission, and playtested that way. I now see that if the user does not interact with the app this data structure was not instantiated, causing the error. It should be fixed now. LMK if it's still not working.
It looks like the data structure that stores the in-game app settings is missing or corrupted in your game.
Is it possible you pressed "continue" in SMCA on your first time playing the game instead of "New Game" or loaded a save from OC? There was an issue with shared caching for the two games, and, while I pushed a quick-and-dirty fix, I still need to add a more comprehensive solution that covers all failure cases.
The upshot is that you could have loaded a save from OC which would not have the app settings data, as that mechanic doesn't exist in OC. If this is the issue starting a New Game should fix it.
Thanks! There are definitely limits to the LLM's capacity to follow instructions, and therefore sometimes the game state is not adequately reflected. I made some tweaks yesterday to improve this, but it will still be <100%. Also, effects tend to persist a few turns after they're turned off due to the influence of the chat history, which I don't necessarily feel is a terrible thing. Easy to rationalize in-universe.
Thanks for the feedback on the feedback on the two games bleeding over. This is something I hadn't considered but is not too surprising given they're built on the same engine and use the same credentials. I need to look into this.
I can't be sure what the issue is on your device without more debugging, unfortunately. It could be caching (try Ctrl F5), no access to https://cdnjs.cloudflare.com, an ad blocker or extension interacting in a way I haven't anticipated, or even an older browser version.
Are you sure your Cohere API key is correct? If you open the developer console (right click inspect in chrome) check for
"POST https://api.cohere.com/v1/tokenize 401 (Unauthorized)" in red. This means your credential doesn't exist in their database.
The narrative progresses but doesn't seem to reflect any changes or you're not getting a response at all? If the former, what sort of prompts are you using. Also, how far are you pushing the sliders? I've only playtested this game myself, and because I made it I'm not going to prompt it the way other players will and won't see failure cases they do, so any feedback would be much appreciated!
On the back end, the spells are implemented primarily through changes to the prompt the LLM receives--spell effects are directly passed as context and spells and effects that alter attractiveness change a latent player/NPC relative challenge rating which determines the instructions to the LLM on how difficult to make the NPC. In practice, Command R+ has limited capacity to follow instructions, particularly in terms of rejecting user commands, so the game is trivial to jailbreak. I have some ideas to resolve this, but they require a more performant model than the current version of Command R+ and I'm waiting for the next Command R+ upgrade or free uncensored LLM API to implement them. For now, the game relies on the honor system--you have to decide not to bullshit the narrator to get the intended experience.
Let me clarify. Temporary modifiers are intrinsically temporary--they're removed whenever you rest (and a few other story specific situations). They are intended to represent short term changes in personality or behavior. Permanent modifiers will persist unless the llm decides they no longer apply to the character due to the narrative. A character with permanent modifier 'shy', for example, might lose that modifier if they become more outgoing during their interactions with you. On the back end, the llm is asked what modifiers it thinks should be added or removed, and can return '-tag' to remove a tag and 'tag' to add one.
Thank you for the feedback. Did the Orgone value go to something like "NaN"?
If you're willing, I'd be grateful for a quick bug report so I can try to replicate and fix the error: https://forms.gle/fSuQCUm1oKpR8DBp6
Modifiers are one of the ways the game changes character personality and behavior. They are generated by AI based on the narration. They are not explicitly limited in number, though hypothetically hundreds of them would break the game. They are also key to several game mechanics: key NPCs that gain the "Loyal" modifier provide passive bonuses and giving an NPC the requisite modifier will resolve Desire side quests.
Click the "Locations" button. This should open a 5x5 grid of locations. Your current location is highlighted yellow. Clicking on a defined location will instantly take you there, highlighting the new location yellow.
Clicking on an "open" location will open a dialogue to name the location. When you submit, a description of the location will be generated and npcs will be procedurally generated if there are currently none. This may take a few seconds. The location's name will change to what you wrote and you can now visit this location by clicking it again, turning the location yellow.
Ultimately, your compliance with this and other terms of the TOS depends on how you, the user, interact with the game. The game is a sandbox and your engagement in erotic scenes--and how pornographic they are--depends on your own choices.
That being said, the following facts may be relevant to you.
1) Despite the CYA boilerplate in the TOS, the Command R+ model is so pornographic that enterprise clients regularly complain that it turns e.g. customer service interactions NSFW. I can provide examples from the official Cohere Discord if you'd like.
2) In practice, it is almost exlcusively used for NSFW roleplay. The model is not performant relative to industry leaders like GPT, Claude, etc. and the only advantage it has is that it is uncensored, unlike leading models. This is a choice Cohere has made to differentiate themselves.
3) You do not need to identify yourself to Cohere to receive an API key. A burner email will work.
4) I have found no evidence of anyone ever being banned for a TOS violation of this sort, or even more obvious and egregious TOS violations such as users creating up to 17 (!) free API accounts for erotic roleplay. I do not suggest you make numerous free accounts to evade their free tier limits as it is certainly against their terms, but mention it to illustrate the degree of enforcement even on issues that are far more salient to them due to the huge compute cost of serving unlimited free API calls to individual users.
Thank you for your helpful comments. I'm working on a browser version to address the first point. Unfortunately, the Highly variable images are an intrinsic limitation of pulling images from gelbooru based on tags--the gelbooru tagging system is very flat so there is no way to connect tags to characters, leading to tag bleed across characters. There also may simply be no image that hits all the tags. I've set it up to prioritize the sex position over character traits in this case. My longterm goal is to have a set of pre-baked scenes and poses the llm choses from based on the narration, then insert characters in, but this is substantially more work. The gelbooru images are a placeholder.