SelenaCore Update: 60x Faster Command Processing with MiniLM Integration

We are excited to announce a major performance breakthrough in how SelenaCore understands and processes your voice commands. By integrating the MiniLM-L6-v2 embedding tier 47cf9a9, we have made our intent router 60 times faster than standard Large Language Models. This means your smart home now reacts almost instantly to your requests without sacrificing precision. In our latest tests, this new sentence classifier achieved an impressive 39 out of 40 accuracy rate on real-world voice data d651714. We also improved how individual modules, such as Media and Presence, register their own command descriptions, ensuring a perfectly synchronized system every time it starts up 861fc80. Additionally, we refined the router's logic with better synonym support and character limits to keep interactions smooth and reliable 51b585e.

What changed:

• Integrated MiniLM-L6-v2 for 60x faster intent routing compared to LLMs.
• Achieved near-perfect accuracy in core command detection.
• Introduced module-owned intent descriptions for cleaner system architecture.
• Enhanced boot-time resync and synonym handling for natural interactions.

#SelenaCore #LocalAI #OfflineAI #SmartHome #EdgeAI #PrivacyFirst #OpenSource