Voice-based speaker ID triggered by YAMNet speech detection. Cosine similarity matching against SQLite enrollment DB. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
