SelenaCore Update: Native Piper TTS Integration and Intelligent Hardware Orchestration

In our latest development cycle, SelenaCore has achieved a major milestone in speech synthesis and local execution efficiency. The primary highlight is the introduction of a native Piper TTS server 7d56c13, providing a portable, high-performance text-to-speech service with a shared model directory and full support for both CPU and GPU acceleration. This shift allows for faster response times and more reliable voice interactions.

To complement the new voice infrastructure, we have implemented automated hardware detection within the LLM orchestration layer. The llamacpp-start.sh script now intelligently identifies the presence of a GPU to optimize model loading d4af6f9. Furthermore, the intent router has been refactored to use UI-configured models directly, replacing the previous static RAM gate with a more flexible availability check 16cf127.

On the stability front, we addressed a critical memory leak in the VLC media handler by ensuring proper release of media objects a028e33. We also enforced language consistency for local LLM providers to prevent mismatched voice responses 02117aa. Comprehensive tests for TTS streaming, GPU detection, and VLC lifecycle management have been added to ensure long-term reliability.

For detailed deployment instructions on the new native piper-server, please refer to the updated documentation.