"To be usable a stim must be present in all devices"

BobRoss235

Hey,

I just wanted to take a moment to say how much I appreciate this website. The work you’ve put into it is fantastic — thank you so much!

I have a question about the note that says:
“To be usable a stim must be present in all devices with the exact same name.”
I’m not entirely sure I understand it correctly, so this is both a clarification request and a feature suggestion in case I do.

My current setup includes a Coyote along with additional toys connected through Intiface and Lovense. I’ve created custom stims specifically for the Intiface devices, but I can’t use them because they aren’t present on the Coyote. In AI mode, the Intiface-only stims are ignored and I get an error that no stims are enabled. I can't add anything to Lovense but actually adding the same stims to Lovense&Coyote doesn’t really make sense either, since I’d prefer individual stims tailored to each device (the standard ones don’t fit my Intiface toys).
I’ve also noticed in debug mode that the AI already selects different stims per device, which makes me think per-device stims could work well.

In short, I’d love it if stims could be device-specific and didn’t need to exist on every device with the exact same name. Each stim already has a description for the AI, which should provide enough context.
Would it be possible to support this?
Thanks again for all your hard work!

KairosJS

BobRoss235 Yes, you understood correctly. To use the stims you crafted for Intiface, you need the "same" stims on the Coyote 3. But in practice only the name has to match, you can create two completely different stims that share a name (just note that the description sent to the LLM will be the one from the first device in your setup).

One thing that may help: this constraint only fully applies to pleasure stims. Pain stims are already exempt for Lovense/Intiface devices, so you don't need to duplicate those.

Here is what is currently sent to the LLM:

<modes>
A mode is a repeating stimulation pattern. Ordered from mildest to strongest within each category.
**Pleasure Modes:**
- \"tickles\": A continuous, steady stimulation pattern at a barely perceptible intensity, intended for very light teasing and frustration.
- \"light_stutter\": A low-intensity, regular staccato pattern alternating between brief maximum pulses and complete pauses, intended for light, rhythmic teasing, and building frustration.
- \"light_pulse\": A low-intensity, regular pulsing pattern that ramps linearly up to a gentle peak and back down to zero, intended for light, gentle teasing, and building frustration.
- \"teasing_pulse\": A medium-intensity, regular pulsing pattern that ramps up and down smoothly, providing pleasant touches and moderate arousal.
- \"teasing_stutter\": A medium-intensity, regular staccato pattern alternating between pleasant pulses and pauses, designed for rhythmic stimulation and moderate arousal.
- \"teasing_continuous\": A continuous, steady stimulation pattern at medium intensity, meant to be clearly felt while maintaining sustained teasing and controlled arousal.
- \"edging_continuous\": A continuous, strong stimulation at high intensity, delivering intense pleasure and pushing toward climax.
- \"edging_pulse\": A high-intensity, rhythmic pulsing pattern that ramps up and down smoothly, designed for intense pleasure and driving toward climax.
- \"edging_wave\": A high-intensity, smooth rising-and-falling irregular pattern, creating intense pleasure and edging toward climax.


Stop: set mode to \"none\".
</modes>

If stims were different from one device to another, it would mean sending a full list for each device. I'm wary of confusing the LLM and making it more prone to hallucination (the context is already large). It's definitely possible though, and I may change my mind in the future with stronger models.

A middle ground I'm considering: keeping a single list of descriptions, and only stating per body area which modes are available there, that would cost almost nothing in context. Good suggestion, I'll keep this on my mind. I'm still afraid that the LLM may start sending stim to the wrong body area, which would probably brick the session.

BobRoss235

Hey,
Thank you so much for the detailed clarification — that really helped me understand the design constraint and the reasoning behind it.
I appreciate the note about pain stims already being exempt for Lovense/Intiface devices, and the practical workaround of just matching names (even if the actual patterns differ). The concern about bloating the LLM context and increasing hallucination risk also makes complete sense, especially with the current setup.

One idea that might strike a middle ground: introduce a limit on the total number of unique stims that can be active in a scenario (e.g. 10 or fewer). Shared stims (same name across devices) would count as one, while device-specific stims would each count separately. That way people who want tailored stims per device could still do so, but only within a reasonable budget that keeps the prompt size manageable.

Either way, I really appreciate you keeping the suggestion on your radar. Thanks again for the thoughtful reply and for all the work you’re putting into this!