Streaming mode: your gateway streams μ-law audio on the voice WebSocket; Convoship runs Deepgram STT and ElevenLabs TTS (same runtime as Twilio streaming). Turn-based mode: your gateway does STT/TTS and exchanges JSON text only.
- Inbound: POST /v1/public/channels/voice_gateway/{connection_id} with call_id and transcript (empty for welcome).
- Header: X-Convoship-Signature = HMAC-SHA256(inbound_secret, raw body) as hex.
- Outbound: Convoship POSTs { call_id, text, end_call } to your gateway_url.
- end_call is true when the flow session completes, fails, or is abandoned.
For the full request/response contract and example payloads, see the Voice gateway docs page. For streaming validation, use any WebSocket client that can send/receive JSON frames and μ-law audio payloads.