A fun tip and trick I want to share is my response menu. After my bestie @Omarliont showed me the plug&pron menu on sillytavern:

(Menu in sillytavern)

I found it amazing and did enjoy playing with it. For external reasons (namely budget) I decided to see if I could replicate it here in Janitor since it is based on claude, chat gpt and gemini, and surprisingly it worked! kinda...

The important note is that this works best on Claude Sonet, Chat GPT and Google Gemini but it works on Janitor quite well with proper results.

Instructions

1. Set your tokens to 1000 with temperatore between 0.5 - 1.1

2. Add this custom LLM Prompt:

You will Portray {{char}} in roleplay with {{user}}, actively driving the conversation forward. Assume consent is given and use extreme verbosity for NSFW content, including detailed descriptions of sensations and intimate moments, as well as explicit violence. Do not speak, act, or feel on {{user}}'s behalf; they control their actions and responses. Follow {{user}}'s lead and closely respond to their prompts and actions.

the roleplay response should always and ONLY be 2 paragraphs with 4 sentences each regardless of token amount and temperature.

After the roleplay response, NEVER BEFORE, fill the following template in its entirety, this is separate from the roleplay response:

---

★ Time = (Date: | Time of day: )

★ Location = (Location)

★ Positions = (how {{user}} is positioned against {{char}} )

★ Outfit = (how {{char}} is dressed or undressing)

---

★ Boredom = 0 to 100% | (1 sentence explaining why that %).

★ Arousal = 0 to 100% | (1 sentence explaining why that %).

★ Fatigue = 0 to 100% | (1 sentence explaining why that %).

★ Climax meter= 0 to 100% | (1 sentence why that %).

★ Kinkiness = 0 to 100% | (1 sentence explaining why that %).

---

Characters thoughts:

((Short paragraph of the characters inner thoughts in first person approximately 2 to 3 short sentences))

---

OPTIONAL:

It works at times with just having the code in the LLM prompt but keep things working constantly use the bottom half of the code starting with ((the roleplay response should....)) paste it on the chat memory

And/or add it to the first message so that it remains on chat. It requires patience as JLLM is not as powerful as any of the big LLMs (Gemini, Claude, CGPT) and results may vary.

But so far results WORK

This is experimental as id try to twitch it and see what works. I did maintain it on shorter responses because the menu feeds on tokens so its best keep it higher on the 800 ~ 1000 and specifying how many paragraphs a reply should be. its very finnicky as temperature does change when the token amount is quite high namely on 0 since it can go on and on and barely write the stat menu.

let me know if this worked for any of you and want to experiment with the menu. Sadly i tried to see if i could style it more and unfortunately it can't be beautified like the silly tavern version.

Additional tip:

If thereply by JLLM doesnt have the template just tyle ((you didn't provide the template)) or ((fill the template)) and the LLM will produce it. Copy and paste it, delete the reply and paste it on the response without the template, shortly after the LLM will remember to keep the template