@happyvertical/browser-ai

Consolidated into smrt-svelte

browser-ai is no longer a standalone package. As of v0.20 it was consolidated into @happyvertical/smrt-svelte, and that consolidation remains the shape on the current v0.29.34 baseline. All browser AI functionality — speech-to-text (STT), text-to-speech (TTS), LLM inference, capability detection, model download progress, and the warm client cache — ships inside @happyvertical/smrt-svelte. The STT/TTS/LLM adapter classes and capability detection live at the bundled @happyvertical/smrt-svelte/browser-ai subpath, while the hooks (useSTT/useTTS/useLLM) and warm-client cache helpers are exported from the package root.

Migration

If you are still on a pre-0.20 import, update to @happyvertical/smrt-svelte:

// Before (v0.19)
import { BrowserAI } from '@happyvertical/browser-ai';

// After (v0.20+, still current on v0.29.34)
import {
  useSTT,
  useTTS,
  useLLM,
  // Warm-client cache helpers are exported from the package root
  getCachedSTT,
  getCachedTTS,
  getCachedLLM,
  getCacheStats,
  clearAllCaches,
} from '@happyvertical/smrt-svelte';
import {
  VoiceInput,
  CapabilityGate,
  DownloadProgress,
} from '@happyvertical/smrt-svelte/browser-ai/svelte';

Available adapters today: STT (browser-speech, whisper-cpp, whisper-wasm), TTS (browser-synthesis), LLM (webllm, transformers-llm). See the smrt-svelte page for the full Provider configuration and preload strategies.

Go to smrt-svelte