This repo contains a fullstack web app for experimenting with the Live Gemini API (using a Gemini API key).
There are two backend deployments:
adk_live_agent: Uses Agent Development Kit (ADK) to build a Live agent. A frontend using NextJS is available inweb-client/gemini_live_cli.py: Uses the genai-processors library (by Google Deepmind) to create an interactive CLI session with Gemini Live API (also supports screen sharing and webcam)
Pre-requisites:
- Python 3.7+
- uv
- Gemini API key obtained in Google AI Studio
Run the following:
uv lock
Create a .env. file and add GEMINI_API_KEY environment variable.
Run the ADK live agent backend:
uv run uvicorn main:app
Run the web client:
cd web-client
npm install
npm run dev
Alternatively, you can use a simpler, Javascript based frontend served directly by the backend.
uv run python gemini_live_cli.py