I often make bots that JLLM can hardly handle, so I'm here to help my dear subs with this issue!

I hardly see any visual guides here, so it's probably better to do it myself so that others don't suffer as much as I did about year ago...

Here, I’ll explain everything as simply as possible. My mom used to say I’m good at explaining things in the simplest possible terms... but the comments are always open, and I love helping people.

To begin with, you create an account on openrouter.ai ! It's not complicated, you can do it EVEN FROM YOUR PHONE!

Next create your own api key and just COPY it, and save somewhere:

Now back to janitor ai..

Few quick steps

Now to our configuration

For Proxy URL: https://openrouter.ai/api/v1/chat/completions

API key: paste your API key that you create on site

Model: use any free one, for example google/gemma-4-31b-it:free or nvidia/llama-nemotron-embed-vl-1b-v2:free

Thats it, now you have less stupid LLM for roleplay.

And btw, if you don't pay openrouter 10$ you probably can't send more than 50 messages daily, so just create 3-4 accounts and make 3-4 configurations with different API keys to bypass this shi

Just keep in mind one simple thing: Prices for AI models change frequently, so if you encounter an error, just look for another free Model on the openrouter website, where you can set the filter to "Free"

By the way, openrouter is greedy, so there are few free models on it, you can search for other sites like HuggingFace and look there. (Proxy URL prob looks like this https://router.huggingface.co/v1/chat/completions)

Otherwise, you can drop 10 bucks into the site's account and use the cool deepseek v4 Flash model for about a year, as I do.