Streaming Vision Agent

This guide provides instructions for setting up and running a streaming vision agent using the Agent Development Kit (ADK).

1. Setup Environment & Install ADK

It is highly recommended to use a Python virtual environment for this project.

Create and Activate Virtual Environment

# Create a virtual environment
python -m venv .venv

Activate the virtual environment

# Create a virtual environment
source .venv/bin/activate

Install Libraries

# install libraries
pip install -r requirements.txt

2. Project Structure

Create the following folder and file structure:

directory_folder/               # Project folder
├── app/                     # Web app folder
│   └── .env                 # Gemini API key file
└── vision_agent/     # Agent logic folder
    ├── __init__.py          # Python package init
    └── agent.py
    └── prompts.py            # Agent definition

3. Set Up the Platform

Choose one of the following platforms :

• Google AI Studio (recommended for local development)
• Google Cloud Vertex AI (recommended one)

Generate your API Key.

Create a .env file inside the app/ folder with the following content:

# Create .env file
GOOGLE_GENAI_USE_VERTEXAI="TRUE"
GOOGLE_CLOUD_PROJECT="aisee-ahlab"
GOOGLE_CLOUD_LOCATION="us-central1"
GOOGLE_API_KEY=xxxxxxxxxxxxxx
RAG_CORPUS='projects/aisee-ahlab/locations/us-central1/ragCorpora/2305843009213693952'

4. Run Agent

Now you're ready to try the agent.

Navigate to the App Directory First, change your current directory to the app/ folder.

# change directories
cd app

Set SSL_CERT_FILE (Required for Voice & Video) This step is necessary for enabling voice and video features. (Running in WSL/Linux is recommended)

# change directories
export SSL_CERT_FILE=$(python -m certifi)

Launch the Dev UI Finally, run the adk web command to start the development user interface.

# launch web
adk web