Move from OpenAI Realtime to Prism

Prism uses the event names and message shapes of the OpenAI Realtime beta interface. So your agent keeps its conversation logic, instructions, and tools when it moves, and most of its code stays as it is. Pick the path that matches how your agent connects today:

Your agent today	What to do
A WebSocket client on your backend	Change four things
LiveKit Agents with the OpenAI plugin	Swap the model
OpenAI’s SDK, WebRTC in the browser, or SIP	Connect from your backend with the Rime SDK, or bring calls in through LiveKit

Move a WebSocket client

The code that sends audio, tool results, cancels, and playback reports stays the same. Change these four things:

Point the client at Prism

Use Prism’s address, and send your Rime API key in the same Authorization header. If you don’t have a key yet, create one.

url = "wss://api.openai.com/v1/realtime?model=gpt-realtime"
headers = {
    "Authorization": f"Bearer {os.environ['OPENAI_API_KEY']}",
    "OpenAI-Beta": "realtime=v1",
}

url = "wss://prism.api.rime.ai/v1/realtime"
headers = {"Authorization": f"Bearer {os.environ['RIME_API_KEY']}"}

Connect from your backend, because Prism has no ephemeral keys for browsers or phones.

Trim the session settings

Keep instructions and tools, and choose a Prism voice. Remove the settings that Prism handles itself, such as transcription, audio formats, and temperature. Prism also finds the turns, so keep only interrupt_response and eagerness in turn_detection:

{
  "type": "session.update",
  "event_id": "evt_setup",
  "session": {
    "voice": "aurora",
    "instructions": "You answer the phone for Juniper Cafe. Call get_opening_hours before you answer a question about opening hours. Answer in one short sentence.",
    "tools": [
      {
        "type": "function",
        "name": "get_opening_hours",
        "description": "Today's opening hours for Juniper Cafe.",
        "parameters": { "type": "object", "properties": {} }
      }
    ],
    "turn_detection": { "interrupt_response": true, "eagerness": "auto" }
  }
}

If Prism refuses a field, the error’s param names it. The WebSocket reference says what happens to each OpenAI field.

Send the caller's audio at 16 kHz

Prism replies at 24 kHz, as OpenAI does, but takes the caller’s audio as PCM16 at 16 kHz, mono. Resample the caller’s audio before you append it, and keep streaming it for the whole call, silence included.

Add a cause to the replies you start

On Prism, every response.create says why it asks for a reply. The most common one is the reply after a tool result:

{ "type": "response.create" }

{
  "type": "response.create",
  "event_id": "evt_continue",
  "response": {
    "metadata": {
      "prism_cause": "tool_continuation",
      "prism_parent_response_id": "resp_0aae3adb2ac71e2c"
    }
  }
}

Set prism_parent_response_id to the response.id of the reply that made the calls. Send it after that reply’s response.done and a conversation.item.created for each result. Typed text and greetings have their own causes, as Typed turns and nudges shows.

Most server events keep their beta names, so your handlers keep working. Read the reply text from response.text.delta instead of response.audio_transcript.delta, and take tool arguments from response.function_call_arguments.done. If your client uses OpenAI’s current event names, such as response.output_audio.delta, map them to the beta names.

Move a LiveKit agent

Prism plugs into LiveKit Agents as a realtime model, the same way OpenAI Realtime does. So the change is the model that you pass to AgentSession:

from livekit.plugins import openai

session = AgentSession(
    llm=openai.realtime.RealtimeModel(voice="marin"),
)

from livekit.plugins import rime

session = AgentSession(
    llm=rime.realtime.RealtimeModel(voice="aurora"),
)

Remove any vad or turn_detection settings, because Prism finds the end of each turn. Your Agent, its instructions, and its @function_tool tools stay the same. Install the Rime plugin as the LiveKit quickstart shows.

Switch to the Rime SDK

If your backend uses OpenAI’s SDK, switch to the Rime SDK in Python, Node.js, Go, or Rust. Its conversation runner plays each reply and stops playback when the caller talks over it, so your client’s playback and interruption code goes away:

async with client.realtime.connect(voice=voice, instructions=instructions) as session:
    await run_conversation(session, source=browser, playback=browser, on_event=show)

The SDK quickstart runs a complete agent.

Check that it works

Make one call. You should see session.updated, then input_audio_buffer.speech_started and speech_stopped while you talk, and then Prism’s reply, ending with a response.done whose status is completed. If an error arrives, its code says what to fix, as Errors and limits explains.

What you can add next

Once your agent runs on Prism, you can:

Get started

Build with Rime

Integrations

Account & updates

Move from OpenAI Realtime to Prism

Move a WebSocket client

Move a LiveKit agent

Switch to the Rime SDK

Check that it works

What you can add next

​Move a WebSocket client

​Move a LiveKit agent

​Switch to the Rime SDK

​Check that it works

​What you can add next

Move a WebSocket client

Move a LiveKit agent

Switch to the Rime SDK

Check that it works

What you can add next