PROXY TUTORIAL!

𝒄𝒊𝒂𝒐, 𝒃𝒆𝒍𝒍𝒂𝒔! 🌹

𝐚 𝐬𝐭𝐞𝐩-𝐛𝐲-𝐬𝐭𝐞𝐩 𝐩𝐫𝐨𝐱𝐲 𝐭𝐮𝐭𝐨𝐫𝐢𝐚𝐥
𝐟𝐨𝐫 𝐧𝐚𝐧𝐨-𝐠𝐩t & 𝐣𝐚𝐧𝐢𝐭𝐨𝐫𝐚𝐢.

________________________

if you follow me, you know i test most of my bots (and make them) with the use of proxies in mind. JLLM is definitely improving (or on its way to improving), but as we all know, it's still on beta. to get the MOST out of my bots (or any others, really), i definitely recommend using a proxy. now, lots of people have told me they don't even know what that means.

first and foremost: most good proxies are NOT free. so let's get the cat out of the bag right away. there ARE some good free proxies out there, but they're usually free only for a time for testing purposes and quickly get taken down after enough people discover them.

there are different websites through which you can get your proxies. for this tutorial, i'll talk about the one i'm using, which is nano-gpt.

so let's get started.

one. what is a proxy?

think of a proxy as a middleman between you and the actual AI model. when you use JanitorAI normally with JLLM, you're using JanitorAI's own language model. when you use a proxy, you're routing your conversation through an external service (nano-gpt) that connects you to different (and often more powerful) AI models like GLM 5, Kimi 2.5, or others.

why does this matter? better models typically mean:

• smarter responses that stay in character

• better memory and consistency throughout long roleplays

• less filtering (depending on the model you choose)

• more creative and nuanced writing

the proxy handles the connection, and JanitorAI just sees it as another API endpoint to send requests to.

two. how much does nano-gpt cost?

nano-gpt offers a few different options:

• [ tier - cost - what you get ]

• [ pay as you go (PAYG) - you decide - add money to your nano-gpt wallet/balance and use any models you want. each attempt at a generated message has a cost based on tokens, lorebooks, model you use, etc. i don't recommend this unless you are more aware of what everything does and costs.]

• [ pro - 8 USD/month - 60 million included input tokens per week, access to most models] (this is the one i use & recommend!)

the free tier is great for testing