AI runtimepackages/ai/voice-realtime
voice-realtime package
Realtime voice loop — input ASR, model stream, output TTS — sub-300ms end-to-end. Bridges OpenAI Realtime + Deepgram + ElevenLabs.
- Stability
- Stable
- Scope
- Global
- Boundary
- packages/ai/voice-realtime
voice-realtime
packages/ai/voice-realtimeAI runtime · tool registry · MCP boundary
Ready
Capability graph
ai/voice-realtimevoice-realtime
Tool policy
Stream UI
Output
Run ledger
- 1Inputpackages/ai/voice-realtime/request
- 2Planmessage reducer -> model turn
- 3Tool policytool schema · auth scope · rate limit
- 4Stream UIevents -> UI state machine
- 5Output@nebutra/agents response envelope
p50 latency
206 ms
events/sec
335/s
providers
5
eval score
91
Usagevoice-realtime.ts
typescriptvoice-realtime.ts
1import { voiceRealtime } from "@nebutra/voice-realtime";
2
3const result = await voiceRealtime.run({
4 tenantId: org.id,
5 // voice-realtime is part of the AI runtime — composable with other AI primitives.
6 input: payload,
7});