jambonz Frame Serializer

​

Overview

​

Installation

uv add "pipecat-ai"

​

Prerequisites

1. A jambonz account or server with an application that runs the listen verb.
2. A public WebSocket URL for your bot, which the verb’s url points to.

​

listen Verb

{
  "verb": "listen",
  "url": "wss://your-bot.example.com/ws",
  "sampleRate": 16000,
  "mixType": "mono",
  "bidirectionalAudio": {
    "enabled": true,
    "streaming": true,
    "sampleRate": 16000
  }
}

• bidirectionalAudio.streaming defaults to false. While it is, audio sent back is ignored and the caller hears nothing.
• sampleRate defaults to 8000. Other allowed values are 16000, 24000, 48000 and 64000.
• bidirectionalAudio.sampleRate has no default. Set it to the same value as sampleRate, or set the serializer’s audio_out_sample_rate to it: the serializer sends audio at the sampleRate from the preamble unless audio_out_sample_rate is set, so a mismatch plays back too fast or too slow.
• mixType must be mono. Audio is dropped with a warning for stereo or mixed.
• Set passDtmf to receive keypad digits as InputDTMFFrame.

​

Configuration

​

InputParams

​

Properties

​

Usage

​

Basic Setup

from pipecat.serializers.jambonz import JambonzFrameSerializer
from pipecat.transports.websocket.fastapi import (
    FastAPIWebsocketParams,
    FastAPIWebsocketTransport,
)

serializer = JambonzFrameSerializer()

transport = FastAPIWebsocketTransport(
    websocket=websocket,
    params=FastAPIWebsocketParams(
        audio_in_enabled=True,
        audio_out_enabled=True,
        serializer=serializer,
    ),
)

​

Notes

• Interruptions: An InterruptionFrame sends killAudio, which flushes audio jambonz is playing or has buffered.
• DTMF support: With passDtmf set on the verb, digits are converted to InputDTMFFrame objects. mark and playDone events are ignored.
• Preamble first: Audio that arrives before the preamble is dropped.