Add Kokoro TTS as a built-in speech engine in Bolt AI 2, alongside OpenAI audio. It enables low-latency, high-quality, local text-to-speech for LLM responses. Motivation Kokoro TTS is fast and lightweight, ideal for real-time responses streaming as LLM outputs tokens. Many users prefer local speech for privacy, offline use, and cost savings, instead of cloud APIs. Making Kokoro a core, selectable backend enhances Bolt AI 2 for multimodal, voice-first workflows. Functionality Kokoro as a backend in audio/TTS settings. Configure model, voice, speed, and response options. Support streaming: send chunks incrementally, start playback early, create real-time speech. Support local process or remote endpoint with URL, API key, request mapping. UI controls: "Play with Kokoro” button, indicating speech status, voice dropdown. Examples Global: set audio provider, endpoint, default voice, speed. Per assistant: enable auto-speak and streaming. Benefits Private local voice for sensitive use. Low-latency, real-time interactions. Cost savings with local compute. Fits power users with local models and self-hosted AI.