Google Developer: Build a Multimodal Live Streaming Agent with ADK

Author: Google for Developers - Bewertung: 23x - Views:184

Want to build AI agents that can see, hear, and react in real-time? 🤔 We kick things off by diving deep into the challenging process of developing live agents directly with Gemini models' Live Streaming API, showing you what it takes to build complex function calling management from the ground up. But what if there was a much simpler way? Then, discover how the Agent Development Kit (ADK) revolutionizes this, drastically simplifying the creation of powerful live agents by expertly handling core components like live queue requests and event management. ✨ Start building! Watch our full video to learn how.



Chapters:

0:00 - Intro

1:09 - Live Agent Architecture - Gemini API

3:38 - Code - Gemini Live API

5:04 - Demo - Gemini Live API

6:45 - Live Agent Architecture - ADK

8:44 - Code - ADK

15:49 - Demo - ADK

18:22 - Recap and resources

20:24 - Outro



Resources:

Access the ADK streaming documentation: https://google.github.io/adk-docs/streaming

Clone and use our code (Live Streaming API): https://github.com/heiko-hotz/project-livewire

Clone and use our code (Live Agents with A2A& MCP): heiko-hotz/project-horizon "