This Tutorial will walk you through:
I also made a Troubleshooting Guide:
Just read the tutorial carefully before doing anything, I have made sure that this is the easiest way to get it working on JanitorAI quickly. Hopefully when you are done with the tutorial you will have much better chats with the bots on JanitorAI.
Cerebras Systems Inc. is an American artificial intelligence company with offices in Sunnyvale, San Diego, Toronto, and Bangalore, India. Cerebras builds computer systems for complex AI deep learning applications. Of course, you’re not here to read all this stuff, here’s the important part: It has extremely fast inference time!
Note:
You can skip this step if you are already signed in.
First, head to Cerebras.AI Sign In Page. Once there, you can sign in using Google or GitHub, but if you don’t want to, you can create an account with your email directly. If it’s your first time there, it will ask you for Your name and What best describes your case. This data will not be shared publicly, so don’t worry.
On signing up for the first time, it will show you a dialog box titled 🚀 Make your first API call!. In this box, you will see your API key, to copy it, just click the ⧉ button. When you are done, you can click the × button on the top right or the blue Continue button on the bottom right.
Important Notes:
Save this key somewhere so you can find it easily.
Also, don’t share it with anyone, think of it like a password. Of course, you can create another key if the default one got leaked, but it’s safer this way.
If this is not your first time to login to your Cerebras.AI account, you can click the ≡ button on the top right and select API Keys. You can then see a list of your API keys, there’s only one by default. To copy your API key again, just click the ⧉ button next to it in the list.
To list the available models, click the ≡ button then select Limits. You will see a table with all the models you can use for free and their respective usage limits. At the time of writing this tutorial, the two best models are llama-3.3-70b and qwen-3-32b, because they have a 65536 context window. If this changes over time, make sure to select a model with a high context window as this is the whole purpose of using a proxy. After determining which model you want to use, copy its name and context window size as we will need them in the next step.
Note:
The Usage Limits that Cerebras.AI put are generously high, but if you do exceed them, you can either wait for the next day or create another account and redo the tutorial (But I don’t recommend it, if they catch you, they will delete your accounts).
Before we start, make sure you have the following:
llama-3.3-70b)65536)To start, head to JanitorAI Login Page (or Sign Up) and login to your account, then, go to the home page select a bot that supports proxies (How to know if a bot supports proxies?). Then, click Chat with [BOT] or Continue latest chat.
After that, once you’re in the chat page, click the ≡ button on the top right and select API Settings. Select the tab named Proxy instead of JanitorLLM Beta. You will then see lots of stuff, I know, I know, just don’t let it freak you out, it’s really simple. Here’s what you need to do:
Default Configuration, then click EditConfiguration Name field, type Cerebras.AI (or anything you want, it’s just a name)Model Name field, enter the ID of the model in the input.Proxy URL field, enter https://api.cerebras.ai/v1/chat/completions.API Key field, enter your Cerebras.AI API key.Custom Prompt field, you can optionally enter a custom Jailbreak prompt if you have one.Save Changes button (NOT the Save Settings button)Save Settings button on the bottom right.Important Note:
If you use JanitorAI in Incognito Mode or similar, you will need to re-enter the API key into the
API Keyinput field for security reasons. You don’t need to reload the page because of this.
Now, you are almost done, just need to change a few more parameters to make it use the new context window. Click the ≡ button again, but this time select Generation Settings. After this, do the following:
Temperature and don’t change it, it can mess things up. (if you accidentally changed it, set it to 0.9).Max tokens to 0.Context Size to the context window size of the model you chose (eg. 65536) and make sure you use the number input box instead of the slider because the number must be accurate.Save button.And Congratulations!! 🎉🎉🎉 you now have a much better model that can actually remember names and events and not mix things up!
To know if a bot supports proxies, click on it to view its page, then scroll down.
If there is Proxy allowed under the bot’s tags, then that’s what we want, but in some bots, you won’t see it under the tags which means that this bot doesn’t allow proxies.
Another way to guarantee a bot has proxies allowed is that its definition isn’t hidden, so if you see a bot with its definition shown that means it allows proxies. About 90% of all bots on JanitorAI are proxy-compatible.
I personally keep all my bots proxy-compatible for the best experience (I make bots that are huge, so it needs a good LLM to run on).
There are several reasons why you may get an error and that depends on the type of the error:
PROXY ERROR 422 means there’s something wrong with the parameters, to solve this issue, make sure that:
This means that either:
Proxy URL or left it empty.This one is kinda obvious, just make sure to put your Cerebras.AI API key in the API Key field. And remember that the key is automatically emptied if you are in Incognito Mode and you restarted the session (closed the browser window and opened another one then signed in to JanitorAI).
This is because you either have the Temperature too low or too high. I recommend 0.9 for normal use and 1.1 if you want to Jailbreak it for uncensored use.
Tutorial by TheQuietWriter.