Google ADK

Sample Google Agent Development Kit (ADK) agents with callable tools. Each agent can be deployed to Vertex AI Agent Engine, and published to Gemini Enterprise (previously Agentspace).

🤖 Agent Summary

Agent	Description
agent_id	Deploys a simple agent with Agent Identity principal (preview)
hello	Greets users and rolls dice using Python functions.
scc	Lists top SCC findings in a project, and remediation guidance for a finding.
web_risk	Detects malware, phishing, and unwanted software URLs using Web Risk API.
security_manager	Multi-agent security manager (includes SCC and Web Risk agents)
secret_agent	Retrieves secrets from Secret Manager bound as environment variables

⚙️ Environment Configuration

Copy .env.example to .env in the respective agent sub-folders, and fill in your configuration:

# Option 1: Gemini API Key (for local development)
GOOGLE_GENAI_USE_VERTEXAI=False
GOOGLE_API_KEY="your-google-api-key"

# Option 2: Vertex AI (for Google Cloud deployment)
GOOGLE_GENAI_USE_VERTEXAI=True
GOOGLE_CLOUD_PROJECT="your-project-id"
GOOGLE_CLOUD_LOCATION="your-location"
GOOGLE_CLOUD_STORAGE_BUCKET="your-storage-bucket"

Note: For Vertex AI deployments, make sure your active Google Cloud credentials have Vertex AI Admin and Storage Admin roles.

🚀 Deploy to Vertex AI Agent Engine

Use poetry to install necessary pre-requisites.

pip install poetry
poetry lock
poetry install --with deployment

Navigate to each agent folder and deploy:

cd hello && python deploy.py
cd ../scc && python deploy.py
cd ../web_risk && python deploy.py

A successful deployment will result in:

✅ Deployed agent: projects/<PROJECT_NUMBER>/locations/<LOCATION>/reasoningEngines/<AGENT_ID>

🌐 Publish to Gemini Enterprise

After deployment, you can publish your registered ADK agents to Gemini Enterprise using the Agent Registration Tool.

1️⃣ Clone the registration tool

git clone https://github.com/VeerMuchandi/agent_registration_tool
cd agent_registration_tool

2️⃣ List deployed agents

python as_registry_client.py list_deployed_agents \
  --project_id "gcp-project-id-from-config" \
  --location "location-from-config"

3️⃣ Create registration configs for each agent. Either use default config.json or --config to specify a different path. Example below

nano config_hello.json

{
    "project_id": "gcp-project-id-from-config",
    "location": "location-from-config", 
    "app_id": "agentspace-app-id",
    "ars_display_name": "agent-display-name-for-registry",
    "description": "agent-description-for-registry",
    "tool_description": "tool-description-for-registry",
    "adk_deployment_id": "reasoning-engine-id-from-agent-engine",
    "auth_id": "oauth-auth-id-if-any",
    "api_location": "agentspace-app-location",
    "re_location": "reasoning-engine-location"
}

Note: If reasoning engine is projects/PROJECT_ID/locations/LOCATION/reasoningEngines/RE_ID, then adk_deployment_id is RE_ID.

4️⃣ Register the agents. If successful, your agents will be visible and callable within Gemini Enterprise.

python as_registry_client.py register_agent --config=config_hello.json
python as_registry_client.py register_agent --config=config_scc.json
python as_registry_client.py register_agent --config=config_webrisk.json