0:00 / 0:00

Building Reliable Speech-to-Speech Voice Agents: Combining Gemini Live + LangChain Deep Agents

Name: Building Reliable Speech-to-Speech Voice Agents: Combining Gemini Live + LangChain Deep Agents
Uploaded: 2026-06-30T17:03:09.627Z
Duration: 8 min 24 s
Channel: LangChain
Description: LangChain demonstrates a voice research agent combining Gemini Live with Deep Agents to achieve fast, natural speech interaction while executing complex research tasks in parallel. The video covers architecture tradeoffs, orchestration patterns, and practical implementation of multi-agent speech systems.

LangChain30 June 2026Watch on YouTube

Part of series

Ep. 9 · LangSmith Engine

LangChain's LangSmith Engine automatiseert het verbeteren van AI-agents door productie-traces om te zetten in geheugen en fixes.

View the series

Description

Caroline di Vittorio, software engineer at LangChain, demos a voice research agent that sidesteps the usual tradeoff between speech-to-speech models and sandwich architectures. The agent uses Gemini Live for natural, low-latency conversation while offloading long-running research tasks to LangChain's Deep Agents, so you get punchy dialogue and accurate, multi-step research at the same time. Chapters: 0:00 The tradeoff every voice agent builder faces 1:00 How sandwich architectures work and where they fall short 1:12 Speech-to-speech models: the benefits and the catch 1:38 Introducing the voice research agent 1:58 Best of both worlds: Gemini Live plus Deep Agents 2:22 Live demo: searching Reddit for cake recipes 3:06 How the two agents are orchestrated under the hood 4:03 Swapping in any Deep Agent, including ones that run for minutes 4:30 Solving the blocking problem with async tool acknowledgement 4:55 Walking through the code 6:45 Reading the trace in LangSmith 7:57 Wrap-up and where to find the code Extra resources - Deep Agents: https://www.langchain.com/deep-agents - LangSmith: https://www.langchain.com/langsmith-platform - Github: https://github.com/langchain-ai/google-adk-realtime-deepagents-example

What you'll learn

You learn how Gemini Live and LangChain Deep Agents are combined for fast, natural speech interactions while handling complex research tasks in parallel
Sandwich architectures require a tradeoff between speed and accuracy, while this system leverages both benefits
Deep Agents can run long-running tasks without blocking the voice dialogue, thanks to async tool acknowledgement

Frequently asked questions

What is the main problem this voice agent system solves?

The system eliminates the choice between speech-to-speech speed and sandwich architecture accuracy. By using Gemini Live for natural dialogue and Deep Agents for complex research tasks in the background, you get both fast responses and accurate, multi-step research.

How does the system prevent long-running tasks from blocking the voice interaction?

It uses async tool acknowledgement, allowing the agent to respond immediately to the user while Deep Agents handle time-consuming research work. This ensures responsive dialogue despite background processing.

What orchestration patterns are used to coordinate the two agents?

The video demonstrates how Gemini Live and Deep Agents are orchestrated under the hood to work together. Deep Agents can be swapped for any other system, providing flexibility for different use cases.