v0.0.1 — Initial Release

CompositeVoice v0.0.1

First release of the CompositeVoice SDK and workspace apps.

SDK

• Event-driven voice pipeline — composable STT → LLM → TTS with typed events
• Providers:
  • STT: NativeSTT (Web Speech API), DeepgramSTT (WebSocket streaming), AssemblyAISTT
  • LLM: AnthropicLLM, OpenAILLM, GroqLLM, GeminiLLM, MistralLLM, WebLLMLLM (in-browser)
  • TTS: NativeTTS (SpeechSynthesis), DeepgramTTS (WebSocket), OpenAITTS, ElevenLabsTTS, CartesiaTTS
• Conversation history with configurable multi-turn memory
• Eager LLM pipeline — speculative generation on Deepgram preflight signals
• Turn-taking strategies — auto, conservative, aggressive, detect
• Server-side proxy adapters for Express, Next.js, and plain Node
• WebSocket reconnection with exponential backoff
• Extensible base classes for custom provider implementations

Workspace Apps

• apps/web — Landing page with version pill and cross-site navigation
• apps/docs — Starlight documentation site with custom header, theme integration, and sidebar
• apps/design — Design system showcases (colors, typography, icons, buttons, alerts, badges, forms, cards, loading, tooltips, tabs, pagination, tables, overlays, banners, code)

UI Package (@lukeocodes/composite-voice-ui)

• Oceanica theme — Teal primary, Stone secondary, Violet accent via color-mix(in oklch)
• light-dark() token system — CSS custom properties that respond to color-scheme
• Preference toggles — Theme, Contrast, Motion, Transparency, Font Size
• Cookie-based preferences — shared across ports (dev) and subdomains (production custom domains)
• Cross-site preference sync — URL param (?cvp=) transfer for different hostnames (Netlify preview)
• Starlight compatibility — data-theme + color-scheme dual-write in ThemeToggle

Examples

Seven example apps demonstrating different provider combinations and SDK features.