Description
google-adk 1.1.1
I use 'gemini-2.5-flash-preview-native-audio-dialog' in audio mode with 'run_live' runner
RunConfig(
speech_config=SpeechConfig(
voice_config=VoiceConfig(prebuilt_voice_config=PrebuiltVoiceConfig(voice_name='Aoede'))
),
response_modalities=['AUDIO'],
output_audio_transcription=types.AudioTranscriptionConfig(),
input_audio_transcription=types.AudioTranscriptionConfig(),
)
This config is needed because I use one agent for audio and text at the same time. When the app switches from audio chat to text chat, the agent's audio transcription continues to be printed in text chat. I have a need to interrupt the output of the model. In audio chat voice interrupt works well. I get an event with interuppted=True.
But when I send a message to the model it starts generating a response and I try to interrupt it with a new message it does not send interrupted=True and keeps sending the response.
Sending messages is done through
LiveRequestQueue().send_content
Going through the chain I see that the message is sent with turn_complete=True, which means that the model should interrupt the response and start replying to a new message, but it doesn't.
Reproduse:
Send message via send_content
When model start answer send another msg