How to pass audio input to an LlmAgent in a SequentialAgent (Gemini 2.5 Flash)

[MartinaBunge98]

I’m building an LlmAgent using the Google ADK and want to pass it an audio file as input. This agent is part of a SequentialAgent pipeline, and I’m using the gemini-2.5-flash model, which supports audio input.

I’m not sure how to correctly pass the audio to the agent. Should it be part of the {var} syntax or is there a specific parameter for audio?

Any guidance or example code for sending audio input to an LlmAgent would be greatly appreciated.

Thanks!

[MartinaBunge98]

I still dont understand how this works whithin a sequential agent. My LlmAgent is the second agent within the SequentialAgent, and as I understand it, it will recieve the path to the audio as the output_key of the previous agent

[boyangsvl]

I will hide the bot answer since it's hallucinating.