@kombifyio/sites-voice
v0.18.5
Published
VoiceProviders for kombify Sites Framework, one package with sub-path exports: /deepgram (STT Nova-3 + TTS Aura-2), /assemblyai (STT-only Universal-3), /speechkit (kombify's own SpeechKit server). Browser-safe short-lived tickets; long-lived keys stay ser
Readme
@kombifyio/sites-voice
Voice provider adapters for kombify Sites Framework. The public release scope is STT-driven Companion control: mint a browser-safe session ticket server-side, stream one utterance to the provider, normalize the final transcript, and submit that text through the normal /api/coating pipeline.
Sub-path exports:
@kombifyio/sites-voice/deepgram— Deepgram STT tickets plus server-side Aura TTS adapter.@kombifyio/sites-voice/assemblyai— AssemblyAI Universal Streaming STT, text-out only.@kombifyio/sites-voice/speechkit— SpeechKit server STT/TTS with server-side wake-word support.@kombifyio/sites-voice/managed— environment-backed provider selection for Node and Cloudflare hosts.
Long-lived provider keys stay on the host. createSession() returns only a short-lived VoiceSessionTicket; Deepgram uses WebSocket subprotocol auth, AssemblyAI uses its short-TTL temp token, and SpeechKit follows the server-provided URL/subprotocol contract.
