I’ve been making bots for myself and decided to start publishing the "finished" ones. Lots of people make great smut bots, so I focus on modern day RPGs and adventure stories. There are tons of zombie RPGs on this site, so you won't see those here unless I have an original idea. I mainly write W4A bots. All of my bot definitions are proxy-enabled and open definitions. Feel free to work from or augment a bot, but I'd appreciate a message back so I can see what you did. Thanks!

*****

Here's how you set up a free proxy:

Most instructions are for DeepSeek, but I recommend setting up Mistral. DeepSeek V3 is objectively better but it’s also much more complex and since it is popular, free servers are almost always overwhelmed or down. Mistral behaves like a better JLLM (in that it is built for writing) but with much higher context. Currently, it is very fast (0.5 sec start times, >200 tokens per second) so it feels as if you almost instantly get the entire response.

Instructions: Click on "using janitor" at the top right. Change the API Settings to "Proxy". Click on "+ Add Configuration".

Configuration Name: Mistral-OpenRouter
Model Name: mistralai/mistral-small-3.2-24b-instruct:free
Proxy URL: https://openrouter.ai/api/v1/chat/completions
API Key: [**ENTER_YOUR_KEY_HERE**]
Custom Prompt: [**LEAVE BLANK UNLESS YOU WANT TO CHANGE**]

You can create an API Key for free here: https://openrouter.ai/settings/keys

Click "Add Configuration" and select your new config. Click "Save Settings". REFRESH THE PAGE BEFORE YOU TEST OR USE YOUR NEW PROXY!

OPTIONAL BUT RECOMMENDED: You get 50 messages/day automatically, which is not a lot. If you have $10 in credits, you get 1000 messages/day, which is PLENTY! Remember, this model is "free", so spending $10 once gets you 1000 messages/day forever (until they change the access rules).

Note 1: There are lots of "free" models on OpenRouter to experiment with. I recommend “google/gemma-3-27b-it:free”, “deepseek/deepseek-chat-v3-0324:free”, and “z-ai/glm-4.5-air:free”. Lots of people like DeepSeek, but the free servers are often overburdened. I especially like the way Gemma writes. If you want JLLM-like writing (but better), use Mistral, since JLLM appears to be a fine-tuned version of an older and tiny Mistral version. Here are your free options: https://openrouter.ai/models?q=free

Note 2: I recommend changing your "Generation Settings" by clicking on the three bars dropdown menu (upper right). I recommend Temperature=0.2, Max tokens=0, and Context Size=32768. Decrease your temperature to get better stability with the character card, increase it to get more creativity. Temperature depends on the model, so models like Gemma prefer temps around 1.1 while DeepSeek is somewhere in between. Max tokens of 0 just doesn’t force cutting off the output. Context Size includes model, history, input, and output.

Note 3: More context slows down generation but gives you more history. Setting to 32k tokens gives you ~24k words for history, or 50 pages of a book: more than plenty for a smut bot. Mistral can do up to 131K but it feels worse the higher you go and your web requests will fail more often. For context, JLLM is stuck at 9k tokens, or less than 10 pages of history and shorter depending on the character card complexity. Either way, consider adding to your "Chat Memory" rather than increasing context beyond 32k.

Note 4: If you get errors, some free providers (Chutes, specifically) will temporarily disable access when traffic is high. This is a limitation of using a free service! You can easily switch back and forth to JLLM or a paid service. Chutes (in particular) will often turn off DeepSeek:free because it is popular but leave on other free models. This can also be nice approach to switch up a bot if it starts to get too repetitive!

Note 5: This is probably a good time to remind you that all online services (especially free ones) could be logged. I do not ever recommend doing anything illegal (or blackmail-able) online but especially not using a free service like Janitor or OpenRouter. Chutes, one of the biggest providers on OpenRouter, is probably logging/mining your chats. If you’re worried about this, currently Venice provides an older Mistral model (still faster and better at writing than JLLM) for free but with safer terms of service: cognitivecomputations/dolphin-mistral-24b-venice-edition:free

Have fun and stay safe!