This update improves the stability, local performance, and hardware acceleration of NOVA LocalAI.
• Enhanced memory stability during long conversations, file analysis, and large code inputs
• Added crash protection to prevent issues caused by pasting too much code or attachments
• Added NVIDIA GPU acceleration with automatic GPU detection
• Added CPU and GPU inference mode switching in settings
• Added a safe CPU fallback if GPU model loading is unavailable
• Improved model loading, context limits, and local processing stability
• Optimized modern interface, chat layout, sidebar behavior, and settings panel
• Added optional code proxy and virtual machine workspace control
NOVA LocalAI runs locally on GGUF models and doesn’t require Ollama.