What is bidi-streaming in Google ADK?

Bidirectional (bidi) streaming allows for simultaneous, low-latency communication where the agent can process and respond while the user is still providing input, supporting natural interruptions.

Can I switch response modalities mid-session?

No, Google ADK requires you to select one response modality (either TEXT or AUDIO) per session; switching requires initializing a new session.

How does session resumption work in this skill?

It provides configurations for SessionResumptionConfig, which allows the agent to automatically reconnect and maintain conversation context across connection timeouts.

Does this skill support both Google AI Studio and Vertex AI?

Yes, it includes patterns for toggling between Google AI Studio and Vertex AI using environment variables without requiring code changes.

What are streaming tools in the context of ADK?

Streaming tools are defined as async generators that yield intermediate results over time, enabling the agent to provide continuous updates for long-running tasks.

ADK Real-Time Streaming Patterns

Name: ADK Real-Time Streaming Patterns
Author: vanman2024

byvanman2024

데이터 과학 및 ML

Configures Google ADK bidirectional streaming to build low-latency, multimodal AI agents with real-time voice and video capabilities.

소개

This skill provides comprehensive implementation patterns for Google's Agent Development Kit (ADK) to enable bidirectional (bidi) streaming between users and AI agents. It covers critical configurations like LiveRequestQueue management, multimodal response handling, and session resumption for connection stability. It is essential for developers building Gemini-powered voice assistants, real-time video analysis tools, and sophisticated multi-agent systems that require seamless context preservation and asynchronous tool execution.

주요 기능

Multimodal support for simultaneous text, audio, and video processing
Bidirectional streaming configuration for real-time AI interactions
Streaming tool patterns using async generators for continuous data output
Multi-agent session transfer for seamless stateful handoffs
0 GitHub stars
Advanced session management with automatic resumption and context compression

사용 사례

Implementing low-latency monitoring tools that stream live data updates
Developing real-time voice and video AI agents using Gemini Live API
Building complex multi-agent workflows that maintain conversation context across sessions

소개

주요 기능

Multimodal support for simultaneous text, audio, and video processing
Bidirectional streaming configuration for real-time AI interactions
Streaming tool patterns using async generators for continuous data output
Multi-agent session transfer for seamless stateful handoffs
0 GitHub stars
Advanced session management with automatic resumption and context compression

사용 사례

Implementing low-latency monitoring tools that stream live data updates
Developing real-time voice and video AI agents using Gemini Live API
Building complex multi-agent workflows that maintain conversation context across sessions