Build a Travel Advisor Agent with Gemini 3.1 Pro

Build an AI Travel Advisor Gemini 3.1 Pro

Amos G.

Published March 11, 2026

Gemini 3.1 Pro has improved reasoning, longer context, and better tool-use capabilities. In this short guide, we'll use the model for LLM orcherstration to create real-time voice AI agent and travel advisor built with Vision Agents.

What You Will Use

Gemini 3.1 Pro: Via Vision Agents plugin.
ElevenLabs: Speech synthesis.
Deepgram: Transcription.
Smart-Turn: Turn detection.
Stream Video: Realtime transport for audio and video.
Vision Agents: Voice/vision framework for Python.

Required Credentials

Google API key (AI Studio)
ElevenLabs API key
Deepgram API key
Stream API key & secret

1. Start With Vision Agents and Plugins Installation

bash

1
2
uv add vision-agents
uv add "vision-agents[getstream, gemini, deepgram, elevenlabs, smart-turn]"

2. Set Your Credentials in a .env

bash

1
2
3
4
5
6
GEMINI_API_KEY=...
ELEVENLABS_API_KEY=...
DEEPGRAM_API_KEY=...
STREAM_API_KEY=...
STREAM_API_SECRET=...
EXAMPLE_BASE_URL=https://demo.visionagents.ai

3. Run this Python Script

python

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
from vision_agents.core import Agent, Runner, User
from vision_agents.core.agents import AgentLauncher
from vision_agents.plugins import deepgram, gemini, getstream, elevenlabs

async def create_agent(**kwargs) -> Agent:
    return Agent(
        edge=getstream.Edge(),
        agent_user=User(name="Assistant", id="agent"),
        instructions="You're a helpful voice/vision AI assistant powered by Gemini 3.1 Pro. Keep replies short and conversational. Be concise and to the point. Always describe what you see in the user's video camera feed.",
        stt=deepgram.STT(eager_turn_detection=True),
        tts=elevenlabs.TTS(),
        llm=gemini.LLM("gemini-3.1-pro-preview"),
    )

async def join_call(agent: Agent, call_type: str, call_id: str, **kwargs) -> None:
    await agent.create_user()
    call = await agent.create_call(call_type, call_id)
    async with agent.join(call):
        await agent.simple_response("Greet the user")
        await agent.finish()

if __name__ == "__main__":
    Runner(AgentLauncher(create_agent=create_agent, join_call=join_call)).cli()

Running this script will allow you join a video call and ask the agent to tell a story or give travel advice. Gemini 3.1 Pro handles the agent's interactions and the pipeline to ensure seamless conversations. Enjoy .

Explore Further

You can experiment with Gemini 3.1 Pro with Vision Agents in several ways. Explore the following resources to learn more.

Python Script.
Vision Agents on GitHub.
Gemini Plugin for Vision Agents.

Scaling WebRTC Video to 100,000 Participants

View Stream's latest Video API benchmark and the architecture that powers performance at scale.

View Video Benchmark