API
Realtime
WebSocket and REST endpoints for real-time voice calls.
The realtime endpoints create voice sessions using the OpenAI Realtime API (WebRTC) and conversational flows.
REST endpoints
Prefix: /api/v1/realtime
| Method | Path | Description |
|---|---|---|
| POST | /sessions | Creates a realtime session for the browser demo. |
| POST | /sessions/end | Ends a session. |
| POST | /tools/execute | Executes a tool during the call. |
| POST | /flow/sessions | Creates a flow-driven session. |
| POST | /flow/advance | Advances a flow-driven session. |
| GET | /flow/sessions/{session_id}/state | Session state. |
| POST | /flow/sessions/{session_id}/end | Ends a flow-driven session. |
| POST | /bookings | Creates a demo booking. |
| POST | /debug-logs | Writes a debug log (local only). |
| POST | /tts | Streams TTS audio. |
WebSocket
| Path | Description |
|---|---|
/api/v1/realtime/tts/ws | Continuous TTS streaming. |
Authentication
Realtime endpoints accept the backoffice session or a public session credential issued when the call is created. For browser calls the session is created by the backoffice and returns session_id and the credential.
Tools during a call
The /tools/execute endpoint allows the model to invoke functions configured on the agent (e.g., transfer, booking, SMS). The request includes session_id, tool_name and arguments.