Feature request: Google ADK integration plugin similar to LangChain

[SebastianBoehler]

Summary

It would be useful to have first-class Google ADK support in livekit-agents, similar in spirit to the existing LangChain/LangGraph integration.

Motivation

Today, if someone wants to use LiveKit Agents together with Google's Agent Development Kit (ADK), they have to hand-roll a custom livekit.agents.llm.LLM adapter / node to bridge the two frameworks.

That is possible, but it is repetitive and awkward compared to the current experience with:

• livekit-plugins-langchain
• provider plugins such as OpenAI / Google / Anthropic

A built-in adapter would make it much easier to reuse existing ADK agents and workflows inside a LiveKit AgentSession.

What support could look like

A dedicated plugin package, for example livekit-plugins-google-adk, that exposes an adapter along the lines of:

• ADKRunnerAdapter(...) or LLMAdapter(...)
• accepts a Google ADK Runner and/or LlmAgent
• translates LiveKit ChatContext into ADK Content
• streams ADK events back into LiveKit llm.LLMStream

Example target UX:

from livekit.agents import Agent, AgentSession
from livekit.plugins import google_adk

session = AgentSession(
    llm=google_adk.LLMAdapter(runner=my_adk_runner)
)

Why this seems to fit the repo well

There is already a precedent for framework-level integration via livekit-plugins-langchain, where a non-provider framework is adapted into LiveKit's llm.LLM interface instead of requiring every user to build the bridge themselves.

Google ADK feels like a similar case.

Useful upstream references

• Google ADK docs: https://google.github.io/adk-docs/
• Google ADK Python repo: https://github.com/google/adk-python

Request

Would you consider adding official Google ADK support, ideally as a plugin package in this monorepo?

Even an initial text-only adapter would already remove a lot of friction for users who want to combine LiveKit's realtime transport/session runtime with ADK agent orchestration.