Hi, we tried Gemini Live API in several ways, what we have noticed with all the supported models: • It never hears the user correctly: tried in English and Dutch • Does tool calling out of nowhere, when it was not requested • User speech is never transcribed whatever we tried, even with the code below in livekit agents 1.3.6:

async def on_session_end(_ctx_: JobContext) -> None:

report = _ctx_.make_session_report()

<http://logger.info|logger.info>(json.dumps(<http://report.to|report.to>_dict(), _indent_=2))

When I do set: _automatic_activity_detection_=types.AutomaticActivityDetection( I get transcriptions like this: noise مثلاً noise منتها это имеет Neu Three different languages, and I was talking Dutch or English Are these by any chance known issues with Gemini Live?