Skip to content

README.md

Latest commit

 

History

History
91 lines (65 loc) · 3.51 KB

README.md

File metadata and controls

91 lines (65 loc) · 3.51 KB

🚀 Deploying GenMedia Izumi Agent to GCP

Welcome to the deployment guide for the GenMedia Izumi Agent! This workspace provides open-source, reference implementations of multimodal creative agents built on Vertex AI.

This guide will walk you through setting up your Google Cloud Platform (GCP) environment and deploying the headless API server using Terraform and Cloud Build.

📋 Prerequisites

Before you begin, ensure you have the following tools installed and authenticated on your local machine:

Tool Purpose Install Link
Google Cloud SDK (gcloud) Interacting with GCP Services Install
Terraform Automated Cloud Infrastructure Install
Python 3.11+ Running setup wrapper scripts Install

Important

Authentication Check: Ensure you have run:

gcloud auth login
gcloud auth application-default login

⚡ Quick Start: 3-Step Deployment

Deploying is a standard sequence of cloning the workspace, provisioning cloud infrastructure, and pushing the server build.

1️⃣ Clone the Repository

Clone this monorepo to your local machine:

git clone https://github.com/GoogleCloudPlatform/genmedia-izumi-agent.git
cd genmedia-izumi-agent

2️⃣ Provision Cloud Infrastructure

We use a lightweight Python wrapper to instantiate Terraform. This script will prompt you for your GCP Project ID and automatically generate your deployment secrets.

python scripts/setup_gcp_project.py --app_env staging

Upon completion, a .env file will be generated in demos/backend/ with your localized cloud bucket bindings.

3️⃣ Deploy to Cloud Run

Once the infrastructure is live, deploy the FastAPI backend container using Google Cloud Build:

./scripts/deploy-to-cloud-run.sh --app_env staging

🎉 Success! The shell will print the final HTTPS endpoint of your serverless AI agent hub.

🛠️ Architecture Deep Dive

If you want to understand what's happening under the hood, this repository treats Infrastructure as Code (IaC) as a primary citizen.

🔐 Infrastructure Provisioning (Terraform)

The resources deployed in deployment/terraform include:

  • APIs: Enables Vertex AI, Cloud Run, Firestore, and Identity-Aware Proxy.
  • Identity & Least Privilege (IAM):
    • Cloud Run Execution SA: Scoped only to Vertex AI + Cloud Storage.
    • Cloud Build SA: Scoped to push images and run deployments.
  • Persistent State:
    • Firestore collections for async job tracking.
    • Cloud Storage Buckets for transient .mp4 stitching buffers.

🚢 The Continuous Build Graph (cloudbuild.yaml)

Triggering ./scripts/deploy-to-cloud-run.sh launches a standard Cloud Build event:

  1. Container Package: It builds the serverless environment pulling from deployment/Dockerfile.
  2. Push to Registry: Pushes the image to Artifact Registry.
  3. Rolling Revision Upgrade: Safely updates the live Cloud Run endpoint to use the new image.

🧹 Tearing Down Resources

To securely remove all Google Cloud resources provisioned for a staging environment and avoid egress billing, run:

python scripts/setup_gcp_project.py --app_env staging --destroy

Caution

Irreversible Deletion: This command uses Terraform to forcibly destroy cloud pools. Database archives and persistent object contents will be deleted forever.