Vagary Voice Developer Platform
Vagary Voice is a real-time voice AI platform. Build applications that hold a live, natural voice conversation — speech-to-text, an LLM turn, and text-to-speech, streamed full-duplex over a single connection — and reach the platform through typed SDKs and a consistent HTTP API.
This site documents the two things a developer needs:
- Client SDKs — the browser / mobile surface your end-user app uses to hold a live voice call.
- Server SDK + API conventions — the typed Python client and the house HTTP conventions (idempotency, cursor pagination, versioning) every capability speaks.
Where to start
| You want to… | Go to |
|---|---|
| Hold a live voice call from a browser or React app | Realtime Client SDK |
| Call platform capabilities from a Python backend | Python SDK |
| Understand the platform's HTTP conventions | API Conventions |
| Get running end-to-end | Getting Started |
The two wires
A live voice call rides one of two transports, both abstracted behind a single VoiceClient:
| Transport | Wire | Direction |
|---|---|---|
| Gateway (default) | voice-gateway /stream | full-duplex: mic in → STT / dialog / TTS → audio out |
| Realtime-media | realtime-media /v1/media/stream | mic in → VAD / turn / barge-in events out (analysis-only) |
The gateway transport is what most applications use — you speak, the agent speaks back. The realtime-media transport is for applications that want the turn-detection and barge-in signals without a synthesized audio return.