Covers model choice, architecture, category mapping, API endpoints, and integration with existing headmic audio pipeline. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>