Voxta 152 Beta: Custom Modules (SDK), Enhanced ExLlamaV2/V3, and All-New Sliders!

Hello everyone,

Voxta Beta 152 is now available. This update focuses on major new features for developers, significant UI improvements, and key fixes for AI models.

Key Additions This release introduces NuGet packages for custom modules, allowing developers to create and integrate their own AI services into Voxta. We've also improved the ExLlamaV2 and ExLlamaV3 models so they work better for both assistant and regular chat interactions. On the UI side, most parameters now use sliders with descriptive badges on hover to make configuration clearer. We've also added a new Text-to-Speech module for Chatterbox, which offers high-quality voices.

☁️ Voxta Cloud Update

• ElevenLabs Price Drop: Prices for ElevenLabs Text-to-Speech on Voxta Cloud have been slashed in half, making it twice as affordable.

✨ Server Updates

• Voxta SDK Modules: Added NuGet packages (Voxta.Sdk.Modules) for custom module development, so you can build your own AI services for Voxta.

• Easier Scenario Sharing: Exporting a scenario now automatically includes the required characters, making them much easier to share.

• Improved ExLlamaV2 and ExLlamaV3: These models now work as intended for assistant and regular chat, with better stop conditions and model suggestions.

• New Chatterbox TTS Module: Added a new module for the Chatterbox Text-to-Speech service.

• Orpheus TTS Sounds More Natural: Refinements like reduced audio gaps and better parameter support (min_p, top_k) make Orpheus voices sound much more natural.

• Custom Voice Embedding: Custom voices are now included in exported character packages.

• Key Fixes:

  • Fixed multiple cases where roleplay text was incorrectly identified as non-roleplay and vice-versa.

  • Fixed an issue preventing chat deletion when a chat contained private memory.

  • Addressed a Florence-2 model loading crash caused by config overrides.

  • Resolved "Cannot access a disposed object" errors in ModuleRuntimeInstanceTicket for better stability.

⚙️ UI Improvements

• Sliders with Badges: Most configuration parameters now use sliders. These have informative badges that appear on hover to explain what each value does, making them easier to adjust.

• Playground Upgrades: The Text-to-Speech and Speech-to-Text playgrounds now have a history feature.

• Redesigned Pages: The entries page for Memory Books and the services configurations page have been redesigned.

• Chat Experience: Added a scroll-to-bottom button in chat views and improved the styling of chat bubbles.

• Confirmation Dialogs: Added inline confirmation dialogs and prompts for key actions.

• Conditional Fields: Added support for conditional and dependent fields in configurations.
  

• Links:

  • How to install Voxta server app: https://youtu.be/1I9VkJ8tTlo

  • How to update Voxta server app: https://youtu.be/5aa7sducwoc

Your feedback is valuable as we continue development. Thank you for your support.