Thanks — I understand you're seeing this error: google.genai.errors.APIError: 1007 None. realtime_input.media_chunks is deprecated. Use audio, video, or text instead. This means the Gemini Live API no longer accepts the old

realtime_input.media_chunks

field. You need to send media via the

audio

,

video

, or

text

fields instead. What you can do next • If you use LiveKit's Google plugin (recommended): configure the AgentSession / RealtimeModel so media is streamed via the plugin's audio/video paths instead of building

media_chunks

yourself. Example pattern in Python:

from livekit.plugins import google

session = AgentSession(
    llm=google.realtime.RealtimeModel(
        voice="Puck",
        temperature=0.8,
        instructions="You are a helpful assistant",
    ),
)
# Stream audio/video through the session plugin rather than setting realtime_input.media_chunks

• If you call the Gemini Live API directly: replace any place that sets

realtime_input.media_chunks

with the appropriate field: ◦ Use

audio

for audio bytes/streams ◦ Use

video

for video frames/streams ◦ Use

text

for text content • Verify package and env: ◦ Make sure you are using a current version of the

livekit-plugins-google

(or google genai client) that supports the new input fields. ◦ Ensure your

GOOGLE_API_KEY

(or other credentials) are set correctly. Quick troubleshooting questions so I can help further 1. Are you using the LiveKit plugin (the

livekit.plugins.google

integration) or calling Gemini Live directly? 2. If you are calling the API directly, can you share (redacted) the snippet that constructs

realtime_input

so I can point out the exact replacement? 3. If using LiveKit, which SDK and version are you running (Python/Node and package version)? If you share which path you're using and a small code excerpt, I’ll give the exact code change to replace

media_chunks

with the correct `audio`/`video`/`text` usage. Sources: Gemini Live API plugin | LiveKit Documentation