Build an AI Travel Advisor Gemini 3.1 Pro
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
uv add vision-agents
uv add "vision-agents[getstream, gemini, deepgram, elevenlabs, smart-turn]"2. Set Your Credentials in a .env
GEMINI_API_KEY=...
ELEVENLABS_API_KEY=...
DEEPGRAM_API_KEY=...
STREAM_API_KEY=...
STREAM_API_SECRET=...
EXAMPLE_BASE_URL=https://demo.visionagents.ai3. Run this Python Script
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.