See full schedule

In this codelab, you will build a full-stack, real-time multimodal AI application with an Agent. You will develop a React frontend to capture live video and audio, and a Python backend using Google's ADK to process this data. By establishing a bi-directional WebSocket connection, you will stream the media to the Gemini Live API, which will interpret hand gestures to execute specific commands—a process known as "Tool Calling.". Author: Christina Lin.


A personal computer is required, no late entries permitted. Facilitators will guide the workshop in Portuguese, but all content and support are also accessible in English.

Building an ADK bi-directional streaming agent [Hosted by Google Cloud]