Vexoria - AI Roleplay Chat (Looking for a Few Testers!) 🎭

[nocare]

Hey all,

Been working on Vexoria (vexoria.me) - an AI roleplay chat platform focused on quality conversations and immersion.

What makes it different:

• Create your own characters - Build custom AI partners with detailed personalities. Keep them private or share with all. Export as a v2 character card.
• Actually remembers context - Hierarchical memory system that compresses old messages into summaries, so conversations can go on for ages without the AI forgetting who you are
• Response validation - Built-in checks prevent hallucinations, keep scenes consistent, and prevent the AI from speaking for you out of turn
• Session persistence - Pick up right where you left off
• Unfiltered - Self-hosted LLM backend, no corporate content moderation (limits apply)
• Handy integration - Optional sync that generates haptic patterns matching the scene

Looking for testers

I’m at the “need real humans to break things” stage. Looking for maybe 5-10 people who:

• Enjoy AI roleplay enough to poke around
• Can tolerate some bugs and provide feedback

New accounts get a few minutes of free time to try it out. After that, it’s Patreon-supported ($10/mo+) - I’m looking for testers who might stick around, not just tire-kickers. Early supporters get to shape what this becomes.

Discord

If you’re interested in testing, please join the Discord - that’s where I’ll be gathering feedback and you can report bugs directly: discord.gg/G3XApj25dS
You’re more than welcome to just come hang out too!

Cheers

Roadmap

Live version: Roadmap

Screenshots

[Amigo6D]

What LLM are you using? And what should we test for?

[nocare]

I’m using a couple of models that felt suitable after testing many. These are pretty light at ~14b. Down the road I would like to fine-tune to dial things in with opt-in data.

I’m mostly looking for feedback on the site and how things work. If something was buggy or just a bad experience I want to know so I can improve it. I really want this to grow into a solid project.

[misuko]

Do testers need to sub to your patreon?

[nocare]

If you sign up on the site you will have some free usage after you verify your email.
It should be enough to get a few message exchanges in and get a feel.

I’m open to rewarding good feedback with a little more usage. I can’t offer all that much at this time though.

[misuko]

Well, I can give some small feedback already. I know nobody usually reads it, but clicking on “Terms of Service” and “Privacy Policy” during registration redirects you to a sign in page, meaning you can’t actually read those before creating an account.

[nocare]

That’s a pretty important one, thank you! DM your username when you have one.

[nocare]

Updates today:

• HSP phase parsing improvement.

  • Haptic scripts should more often have multiple varying parts per message

• HSP generation point-speed reduction

• Removed manual trigger in haptic dropdown

  • Confusing ux

• /terms & /privacy are now visible without login

• Added extra content filtering

• Fixed character name generation (max tokens)

• Fixed portions of HSP scripts overlapping itself

• Improved HSP classification / message-fit

Thanks to everyone who did take time to check things out today. Even if you didn’t say anything, you still helped. A few of you will be getting a little extra usage.

[nocare]

Here are updates from the last couple of days for anyone not in the discord. Thanks to all who have been testing and giving feedback.

• Improved device timeouts
– Handy connections now auto-disconnect if you close the site (immediate) or leave it open in another tab (after 10 minutes).

• Better device isolation
– Fixed an issue where a single chat generation could accidentally send signals to multiple users’ devices.

• Smoother updates (less interruption)
– Updates are now rolled out more gracefully: You’ll only see a refresh popup if an update affects your session. A short grace period lets in-flight messages finish before shutdown. This mainly applies if you’re actively chatting or had a tab open in the background

• Mobile scrolling fixed
– Resolved an issue where the character & user profile pages wouldn’t scroll properly on mobile

• Expanded character tags
– Characters with more than 5 tags now display correctly

• Creator page interaction fixed
– Fixed a bug that prevented clicking or interacting with elements on the creator page

• Improved creator text areas
– Text fields now auto-expand as you type
– Height can be adjusted by dragging

• Mobile responsiveness pass
– General layout and usability improvements across mobile devices

• Responses Now Stream
– AI replies now stream into the chat instead of appearing all at once, making conversations feel much faster and more responsive. To support this, some validation logic has been temporarily disabled. This cuts generation time roughly in half (sometimes more), but may occasionally affect response quality. I’ll be carefully reintroducing quality checks over time to avoid sacrificing speed.
– As a bonus, the speed improvements also give you a bit more message mileage.

• Chat & Input Improvements
– The chat edit bubble has been scaled up for better visibility, no longer forces the page to scroll, and now preserves markdown while editing. Press Esc to cancel edits at any time.
– The chat input now expands as you type, supports Shift + Enter for new lines, and can be manually resized by dragging the bottom-right corner.

• Handy Improvements
– Fixed Handy menu display issues and added support for Handy FW3 . FW3 and FW4 devices are now detected automatically, with playback correctly interfacing based on firmware.
– Automatic Handy connection has also been improved.

• Prevented double message output
– When the AI responds, you should no longer see a second copy of the message appear after typewriting finishes.

• Enhanced typewriting
– When the text is printing you should see less bounces and color flashes

[KingCobraKush]

Heyhey, nice concept. Am trying it out for the moment.

A question though - what are the steps to get a character to ‘be created’? I can fill out the fields in advanced mode and export the json, but there doesn’t seem to be a button to commit the changes into a file the interface can use to actually chat.

Am I missing something?

using Firefox-latest on Windows 10.

[whyareweheretosuffer]

Using latest Firefox. Sometimes chats dont work, or dont send through at all.
Needing to be re-sent or reconfigured over and over until the monthly generation is all gone.

[nocare]

Hey, thanks for check it out. I plan to upgrade that feature soon. Right now it saves a copy to the backend. I so I can manually review and add them on request. You are able to export it for use in something like sillytavern right now.

I will be setting it up so you can have public/private characters after review. I’m just focused on improving chat quality at the moment.

[nocare]

Thanks for checking it out. Can you dm your username? I’ll dig into this further.

[KingCobraKush]

Right on. Thank you for answering.

Not sure what you may want tested further but I have a setup where I can freetest anything you might want.

Let me know if you want a specific thing focused on, and I’m happy to try and help.

[nocare]

I appreciate that and will keep ya in mind. I ran through a ton of UX issues and am currently just focused on getting to a nicer stage with the response quality.

[nocare]

• Character Creation Expansion
– You can now create private or public characters and chat with them.
– There is an approval process in place, so you will have to wait a little bit.
– You can edit and test your own characters. Changes won’t be public until approved.
– Characters now display usage stats

[fksm3ji]

TCode support with webserial?

[nocare]

Not at this time. I have no way to really test. I could emulate and/or test the connection with an esp but that’s about it.

I’ll be adding a buttplug library integration before something like that which will broaden supported devices. I can explore it further after that.

[nocare]

Hey all, I’ve added a simple roadmap to the main post. This will have some of my goals/ideas and hopefully serve as something others can contribute to. I can work on what I want to work on, but if most others want to see something else, I can steer in that direction. As long as it makes sense.

Roadmap

[Flufkoo (Fkoo)]

This looks quite promising! I’m on mobile so I’ll be sure to check it out. Definitely adding intiface support will have such availability for everyone. I hope this grows! Is it going to be trial based every time?