Realtime Audio

Overview

The Realtime Audio API provides session-based voice interaction. A session defines the model, voice, modalities, and turn-detection behavior. Clients connect to a session over WebSocket and stream audio or text in real time.

Base URL

https://api.allternit.com/v1

For local development:

http://127.0.0.1:8013/v1

Endpoints

Method	Path	Description
`POST`	`/v1/realtime/sessions`	Create a realtime session
`GET`	`/v1/realtime/sessions`	List realtime sessions
`GET`	`/v1/realtime/sessions/:id`	Retrieve a session
`DELETE`	`/v1/realtime/sessions/:id`	Delete a session
`WS`	`/v1/realtime/sessions/:id/ws`	WebSocket connection for the session

Authentication

REST endpoints require a virtual key in the Authorization header. The WebSocket endpoint accepts the same key as a query parameter or header, depending on client support. See Authentication.

Create a session

Request body

Field	Type	Required	Description
`model`	string	no	`allternit-realtime-1` or `allternit-realtime-1-mini`. Defaults to `allternit-realtime-1`.
`modalities`	string[]	no	`text` and/or `audio`. Defaults to `["text", "audio"]`.
`voice`	string	no	Voice id. Defaults to `alloy`.
`instructions`	string	no	System instructions for the session.
`temperature`	number	no	Sampling temperature. Defaults to `0.8`.
`max_response_output_tokens`	integer	no	Maximum output tokens per response.
`input_audio_format`	string	no	Audio format for input.
`output_audio_format`	string	no	Audio format for output.
`turn_detection`	object	no	Turn-detection configuration.

Voices

Voice	Description
`alloy`	Neutral, balanced
`echo`	Warm, calm
`fable`	British, story-like
`onyx`	Deep, authoritative
`nova`	Bright, energetic
`shimmer`	Clear, optimistic

Turn detection

Field	Type	Description
`type`	string	Turn-detection strategy.
`threshold`	number	Activation threshold.
`prefix_padding_ms`	integer	Audio padding before the detected speech.
`silence_duration_ms`	integer	Silence required to end a turn.

Example request

curl -X POST https://api.allternit.com/v1/realtime/sessions \
  -H "Authorization: Bearer $ALLTERNIT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "allternit-realtime-1",
    "modalities": ["text", "audio"],
    "voice": "alloy",
    "instructions": "You are a helpful assistant. Keep answers concise.",
    "temperature": 0.8,
    "turn_detection": {
      "type": "server_vad",
      "threshold": 0.5,
      "prefix_padding_ms": 300,
      "silence_duration_ms": 500
    }
  }'

Example response

{
  "id": "sess_abc123",
  "object": "realtime.session",
  "model": "allternit-realtime-1",
  "modalities": ["text", "audio"],
  "voice": "alloy",
  "instructions": "You are a helpful assistant. Keep answers concise.",
  "temperature": 0.8,
  "status": "active",
  "created_at": 1723500000
}

List sessions

curl https://api.allternit.com/v1/realtime/sessions \
  -H "Authorization: Bearer $ALLTERNIT_API_KEY"

{
  "object": "list",
  "data": [
    {
      "id": "sess_abc123",
      "object": "realtime.session",
      "model": "allternit-realtime-1",
      "modalities": ["text", "audio"],
      "voice": "alloy",
      "status": "active",
      "created_at": 1723500000
    }
  ]
}

Retrieve a session

curl https://api.allternit.com/v1/realtime/sessions/sess_abc123 \
  -H "Authorization: Bearer $ALLTERNIT_API_KEY"

Delete a session

curl -X DELETE https://api.allternit.com/v1/realtime/sessions/sess_abc123 \
  -H "Authorization: Bearer $ALLTERNIT_API_KEY"

{
  "id": "sess_abc123",
  "object": "realtime.session.deleted",
  "deleted": true
}

WebSocket events

Connect to the session WebSocket and exchange events:

Event	Direction	Description
`session.created`	server → client	Sent when the session is ready.
`input_audio_buffer.append`	client → server	Stream raw audio chunks.
`input_audio_buffer.commit`	client → server	Signal end of user audio.
`response.created`	client → server	Ask the model to generate a response.
`response.done`	server → client	Response generation finished.
`error`	server → client	Session or transport error.

Example WebSocket connection

const ws = new WebSocket(
  'wss://api.allternit.com/v1/realtime/sessions/sess_abc123/ws'
);

ws.onopen = () => {
  ws.send(JSON.stringify({
    type: 'input_audio_buffer.append',
    audio: base64AudioChunk
  }));
};

ws.onmessage = (event) => {
  const msg = JSON.parse(event.data);
  console.log(msg.type, msg);
};

Error codes

HTTP status	Code	Meaning
400	`allternit.invalid_request`	Invalid `model`, `modalities`, `voice`, or `temperature`.
401	`allternit.authentication_failed`	Missing or invalid API key.
404	`allternit.invalid_request`	Session not found.
429	`allternit.rate_limited`	Per-key rate limit exceeded.
429	`allternit.budget_exceeded`	Monthly key or tenant budget exhausted.
500	`allternit.internal_error`	Internal gateway error.
503	`allternit.upstream_error`	Session store unavailable.

Chat Completions — text-based conversational AI
Files — store audio files for offline processing

Overview

LLM Gateway

Agents & Runtime

Work Queue & Deployments

Files, Webhooks & Sandbox

Tools & Connectors

Collaboration & Workflows

Infrastructure & Platform

Canvases

Onboarding

A Labs

Playground

Users & Profiles

Overview

Base URL

Endpoints

Authentication

Create a session

Request body

Voices

Turn detection

Example request

Example response

List sessions

Retrieve a session

Delete a session

WebSocket events

Example WebSocket connection

Error codes

​Overview

​Base URL

​Endpoints

​Authentication

​Create a session

​Request body

​Voices

​Turn detection

​Example request

​Example response

​List sessions

​Retrieve a session

​Delete a session

​WebSocket events

​Example WebSocket connection

​Error codes

​Related pages

Overview

Base URL

Endpoints

Authentication

Create a session

Request body

Voices

Turn detection

Example request

Example response

List sessions

Retrieve a session

Delete a session

WebSocket events

Example WebSocket connection

Error codes

Related pages