Here's what's new in this update:

• New LLM Roster: We've updated the built-in LLM download list with some of the best new models available:
  • Llama 3.1 8B Instruct (The smartest 8B model right now)
  • DeepSeek R1 Qwen 7B (Currently the absolute best reasoning model in its weight class)
  • Llama 3.2 1B Instruct (A tiny 1GB model that is incredibly fast, perfect for background NPCs)
  • Phi-3.5 Mini Instruct (Microsoft's highly upgraded mini model)
  • Llama 3.2 11B Vision (A new multimodal vision projection model)
• Chat History: Added a new Import/Export option so you can easily save and load your AI Agents' chat histories.
• Memory Management: The MCP Server now features a complete memory and history management system, allowing you to directly add, edit, or delete the AI's memories!

• New Game Added: Welcome Pixel to the Arcade Center lineup!
• Global Leaderboards: Added a global high score system to Tetro 3D! The Top 100 players are now visible live on the Biohazard-Corp Website.
  
  [NOTE] Test Run: This is an early test of the server infrastructure. Scores will be wiped once the system is fully implemented and finalized.
• Alien Mania Fix: Fixed a bug where the player's movement controls were accidentally inverted.

• Upscaling: Added official support for AMD FSR 4, giving users with AMD hardware a great boost to framerates and image quality.

Here's what's new in this update:

• Search Reliability: Fixed a critical bug where the AI web search tool would occasionally fail to return results.
• Custom HTML Crawler: To fix the search issues once and for all, we've integrated a custom, lightweight HTML crawler. It now performs genuine searches and directly extracts titles, snippets, and URLs from webpages—no API keys required!
• Search Providers: The default search engine is now DuckDuckGo. We are planning to add more selectable providers in the near future, including:
  • Brave Search (High-quality results)
  • Serper (Google-powered results)
  • Tavily (AI-optimized results)

• Audio Queuing: Added a new audio queue component designed to properly manage and playback CoquiTTS procedural sound waves without clipping or overlap.
• Settings Sync: Fixed a bug in the audio settings menu where tweaking the volume values wouldn't push the changes to the active audio component.

• Upscaling: Officially added support for NVIDIA DLSS 4.5 for improved framerates and image quality!
• UI Clarity: Added a handy tooltip to the global Fullscreen button, so it is always clear exactly what display mode (Fullscreen, Windowed, etc.) is currently active.

• High Scores: Designed and added a brand-new, stylish "Enter High Score" widget for Tetro 3D to celebrate your best runs.

Here's what's new in this update:

• Engine Upgrade: Successfully updated the project to Unreal Engine 5.8.2.
• Display Settings: Added a new option to the main menu to easily set your preferred screen mode (Fullscreen, Windowed Fullscreen, and Windowed). You can also quickly toggle this using the F11 key.
• Patch Notes: Added a new "Spectra Update Info" section directly into the Support Widget so you can read these notes in-game.

• Readability: Overhauled the chat response system formatting for much clearer and easier reading.
• Quality of Life: The MCP Chat widget now supports quick-sending! Pressing Enter will immediately activate the Ask button. To create a new line in your message, simply use Shift + Enter.

• Quick Search: The Audius search widget now reacts to the Enter key to instantly start searching for tracks.
• UI Fixes: Fixed a bug in the play widget that was causing it to have no scrolling limit.
• Visualizer: Re-added the Niagara Audio Frequency Analyzer for music visualization.

• UI Polish: Updated the Tetro 3D High Scores widget with a sleek, more stylish new look.
• Work in Progress: Recreated the "PiDog" companion and re-added the "Biohazard Space Shooter" game. (Note: Both of these are currently in development and not fully ready to play just yet!)

Here's what's new in this update:

• Fixed a bug where the test music started playing when opening the Music Festival Level.

• Removed all Open level by name references (deprecated).
• Fixed a bug where the arcade games would not open the next level.

Here's what's new in this update:

• Prompt Formatting: Fixed a bug where internal model tokens like </start_of_turn> and <end_of_turn> were visible in the chat and incorrectly sent to the Text-to-Speech (TTS) engine.
• Audio Control: Added a new Stop TTS button, allowing you to instantly interrupt audio generation mid-sentence.

• System Overhaul: Fixed numerous underlying bugs in the Arcade Center framework.
• Games Returned: The classic roster is back in the menu! You'll now see Space3D, Proxima, GridLock, Alien Mania, Orb Mania, and PizzaMan. (Note: These games are visible again but are currently works-in-progress and not fully playable yet).

• Engine Stability: Completely replaced the experimental Audio Niagara scratch module with proper, native C++ code. This resolves a critical issue that was causing the game to crash in Unreal Engine version 5.8.

• UI Polish: Updated and refined the layout of the NavCore settings widget to improve overall visibility and ease of use.

Here's what's new in this update:

• Fixed a bug where the AI didn't see the EOS (end of sentence) token.
• Added Gemma to the model chat template.

• Fixed missing localization in Tetro3D.

Here's what's new in this update:

• Prompt Optimization: Fixed a speed bug related to pre-resolving the {UserName}, {AIName}, and {Mood} variables inside the System Prompt, resulting in faster processing.

• User Identity: Fixed an issue with the displayed username. The user send message bubble will now correctly show your chosen username.

• Configuration Loading: Fixed a race condition (loading too fast!) where changed settings would occasionally fail to load correctly in the packaged game.

• Menu Fixes: Resolved a bug where the Tetro 3D option button was unresponsive.

Here's what's new in this update to refine the Spectra experience:

• Contextual Naming: The AI now correctly utilizes your set user name or user ID when addressing you. (Hint: You can also just tell the AI your name in chat, and it will remember!)
• Settings Visibility: Fixed a bug that was preventing the Personality section from displaying correctly inside the AI settings widget.

• Widget Localization: Updated missing localization strings for the MCP Chat widgets, ensuring better language support.
• UI Fixes: Fixed the display of the user message background image and resolved issues with the "Ask" button in the MCP Chat.

• Tooltips: Corrected the tooltip on the "Search Music" button inside the Music Hub widget for better clarity.

• Navigation: Fixed a bug where the scroll buttons inside the play widget were unresponsive.

Working hard to fix bugs and improving the core experience. Here's what's new in this update:

• Player Stability: Resolved issues with the music player search functionality, ensuring smooth track discovery. Fixed bugs related to opening and closing the player interface and getting stuck.

• Ticketing & Communications: Addressed critical bugs within the ticket system. The issue preventing support emails from sending properly has been fully resolved.

• Font Rendering: Partially fixed global font rendering problems. Most languages now display correctly. (Note: We are still actively tracking and working on resolving lingering font issues specific to Hebrew and Thai).

Welcome to the foundational Early Access release of Spectra! This update represents a complete architectural rewrite, establishing the framework for a decentralized, 100% offline, UE5 intelligence ecosystem.

• Fully Localized Pipeline: Ripped out cloud dependencies. Inference runs entirely locally via ONNX and llama.cpp.
• Modular GGUF Support: Swap in custom quantized GGUF Large Language Models to optimize VRAM allocation.
• Hardware Acceleration: Optimized ML stack leveraging Nvidia CUDA Toolkit 13.2+ for parallel compute.
• Groq API Bridge: Optional high-speed bridge for users lacking local GPUs.

• Neural TTS & STT: Integrated XTTS-v2 for high-fidelity Text-to-Speech and Whisper for local microphone transcription.
• Dynamic Lip-Sync: Sub-100ms real-time audio bridge drives UE5 MetaHuman facial control rigs, converting TTS audio directly to visemes.

• MCP Server Bridge: Custom server ensuring high-speed, secure communication between our HTML/JS UI (UE5 UWebBrowser) and the C++ backend.
• Standalone Executable: Python environments are now fully bundled. No pip installs or manual Python configuration required to run Spectra.
• Mythos Hive (Alpha): Network foundations laid for the upcoming distributed, TCP/IP-based framework to crowd-train the Spectra Mythos model.

• Spectra Neural Interface: Redesigned lightweight, CSS-driven WebBrowser overlay with seamless C++ data binding.
• Music Core & Arcade: Integrated local audio streaming (45 genres, 1M+ tracks) and shipped the initial 3D Arcade featuring Tetro 3D with local high scores.