Voxta v135 Release: Vision Support, Llama.cpp Integration, Enhanced Memory, and More!

🎄✨ Ho, ho, ho! 🎅 Christmas is just around the corner, and we’ve got something amazing for you! ✨🎄

Introducing Voxta Server v1.0.0-beta.135—packed with exciting new features, enhancements, and fixes to make Voxta even more fun to use. Let’s unwrap this gift together:

🎁 Llama.cpp
Voxta now comes with built-in support for Llama.cpp, fully compatible with GGUF models. With features like mirostat, dynamic temperature, XTC, and logit bias, it’s got you covered—even if you’re running on a weaker GPU or just using your CPU and RAM.

🎁 Vision
We now have vision support, so your characters can not only talk to you but also see you and analyze your screen in real-time!

• Continuous Vision: Analyze visual data every few seconds! Be mindful, though—this feature can use up your Voxta Cloud credits quickly.

• Prompted Vision: Activate analysis with keywords like “Look,” “Watch,” or “See.”

• Vision Playground: Test snapshots from your camera or desktop and watch as vision models analyze them in real-time.

New Vision Model in Voxta Cloud: Meet Gemini Flash. It’s very fast and precise. Combining that with long-term memories, you enter a whole new territory where AI characters can not only talk to you but also see you, remember what you show them 😈, how you look, and more.

Prefer self-hosting? Thanks a LOT to @aios for his contributions! You can now run Florence-2 vision models locally:

• Florence-2 Base Model: Needs ~5GB of VRAM.

• Florence-2 Large Model: For more accuracy, requiring ~10GB of VRAM.

🎁 Voxta Packages
It's now easier than ever to manage characters and scenarios:

• Everything is exported in the handy .voxpkg format.

• When copying a character or scenario, all associated assets are copied too—perfect for version control or complex projects.

🎁 Improved Long-Term Memory
Voxta’s memory system is now more powerful, with better recall and organization. Your characters will remember everything—so if you’ve been naughty this year, don’t think you can hide it. Santa’s not the only one keeping track!

✨ And that’s just the beginning! ✨

🎁 Tips to Get Started
Before diving into these features, here are a few quick tips to help you get started:

Enable Vision:

• Navigate to your Voxta folder, open appsettings.json, and set "EnableVision": true.

• Reload Voxta, and you’ll find Vision in the Services menu.

Set Up Desktop Capture:

• Install the Windows SDK module:

  • Go to Manage Modules, click Add Service, find Windows SDK, and install it.

• Configure your desired monitor.

• Install the Vision Module:

  • For Continuous/Prompted vision go to Manage Modules, click Add Service, find Vision, and install it.

Switch Between Webcam and Desktop Capture:

• Open the Vision Capture dropdown in the Services menu panel on the right.

• Look for the Use Browser Webcam checkbox:

  • Checked: Enables Webcam capture.

  • Unchecked: Enables Desktop capture via Windows SDK.

You must set your character to Companion or Assistant type and enable the Vision flag in Augmentation:

• Open the Configuration tab in the Edit Character menu.

• Set the chat style to Companion/Assistant.

• Select Vision in the Augmentation dropdown.

• Finally, toggle on the Vision Support option.

So, make sure to back up your Voxta.db and start exploring the latest version today. Thank you for your support—your feedback is the best Christmas gift we could ask for!

🎄 Happy Holidays! 🎄