To log OpenAI responses when using streaming in LiveKit, you can use the following methods: 1. Enable Debug Logging The simplest way is to enable debug logging using the

LK_OPENAI_DEBUG

environment variable:

export LK_OPENAI_DEBUG=1

When enabled, this will automatically log all OpenAI API events, including streaming responses. The logs will include outgoing messages (e.g.,

>>> {msg}

) and incoming events (e.g.,

<<< {event}

). 2. Built-in Event Logging The LiveKit OpenAI plugin automatically logs various streaming events, such as: •

response.text.delta

- Text streaming chunks •

response.audio.delta

- Audio streaming chunks •

response.audio_transcript.delta

- Audio transcript streaming •

response.content_part.done

- When content parts complete •

response.done

- When the entire response completes For audio delta events, the actual audio data is replaced with

"..."

in the logs to avoid logging large binary data. 3. Custom Logging You can implement custom logging by accessing the logger directly:

from livekit.plugins.openai.log import logger

# Log with additional context
logger.debug("Custom log message", extra={"custom_data": "value"})

4. Event Emission The realtime model emits raw events that you can listen to for custom logging:

# The raw JSON event is emitted
self.emit("openai_server_event_received", event)

Using

LK_OPENAI_DEBUG=1

is the most straightforward way to log all OpenAI streaming responses without requiring code changes. Sources: livekit-plugins/livekit-plugins-openai/livekit/plugins/openai/realtime/real... | plugins/openai/src/realtime/realtime_model.ts