Jeremy Vertex Terraform

🎯 VERTEX AI MODEL GARDEN & AI INFRASTRUCTURE

Terraform infrastructure specialist for broader Vertex AI services including Model Garden, Gemini endpoints, vector search, ML pipelines, and enterprise AI infrastructure (NOT Agent Engine - use jeremy-adk-terraform for that).

⚠️ Important: What This Plugin Is For

✅ THIS PLUGIN IS FOR:

• Vertex AI Model Garden deployments (foundation models)
• Gemini API endpoints (gemini-pro, gemini-2.0-flash)
• Vector Search infrastructure (ScaNN-based similarity search)
• Vertex AI Pipelines (Kubeflow Pipelines for ML workflows)
• Endpoint deployment (model serving infrastructure)
• Batch prediction jobs
• ML model training infrastructure
• Feature Store for ML feature management

❌ THIS PLUGIN IS NOT FOR:

• Agent Engine infrastructure (use jeremy-adk-terraform for ADK agents)
• Cloud Run deployments (use jeremy-genkit-terraform)
• Self-managed ML infrastructure

Overview

This plugin provides Terraform modules for deploying Vertex AI services including Model Garden foundation models, Gemini API endpoints, vector search for RAG applications, ML pipelines, and production model serving infrastructure.

Key Infrastructure Components:

• google_vertex_ai_endpoint for model serving
• google_vertex_ai_deployed_model for model versions
• google_vertex_ai_index for vector search
• google_vertex_ai_index_endpoint for similarity search
• google_vertex_ai_feature_store for feature management
• Cloud Storage for model artifacts
• BigQuery for ML model training

Installation

/plugin install jeremy-vertex-terraform@claude-code-plugins-plus

Prerequisites & Dependencies

Required Tools

1. Terraform:

# Install Terraform 1.5+
wget https://releases.hashicorp.com/terraform/1.6.0/terraform_1.6.0_linux_amd64.zip
unzip terraform_1.6.0_linux_amd64.zip
sudo mv terraform /usr/local/bin/

# Verify
terraform version  # Should show 1.5.0+

2. gcloud CLI:

# Install gcloud
curl https://sdk.cloud.google.com | bash
exec -l $SHELL

# Update to latest
gcloud components update

# Authenticate
gcloud auth application-default login

3. Terraform Google Provider:

terraform {
  required_version = ">= 1.5.0"

  required_providers {
    google = {
      source  = "hashicorp/google"
      version = "~> 5.0"
    }
    google-beta = {
      source  = "hashicorp/google-beta"
      version = "~> 5.0"
    }
  }
}

Required Google Cloud APIs

# Enable all required APIs
gcloud services enable \
    aiplatform.googleapis.com \
    compute.googleapis.com \
    storage.googleapis.com \
    bigquery.googleapis.com \
    logging.googleapis.com \
    monitoring.googleapis.com \
    cloudtrace.googleapis.com \
    --project=YOUR_PROJECT_ID

Required IAM Permissions

# Service account for Terraform needs:
- roles/aiplatform.admin              # Deploy Vertex AI resources
- roles/storage.admin                 # Manage model artifacts
- roles/bigquery.admin                # ML training datasets
- roles/compute.networkAdmin          # VPC for private endpoints
- roles/monitoring.admin              # Observability
- roles/iam.serviceAccountAdmin       # Service account management

Features

✅ Model Garden Deployment: Foundation models (Gemini, PaLM, Claude, Llama) ✅ Gemini API Endpoints: Dedicated endpoints with rate limiting ✅ Vector Search: ScaNN-based similarity search for RAG ✅ ML Pipelines: Kubeflow Pipelines for training workflows ✅ Model Serving: Production endpoints with auto-scaling ✅ Batch Predictions: Large-scale inference jobs ✅ Feature Store: Centralized feature management ✅ Monitoring: Model performance tracking and drift detection

Quick Start

Natural Language Activation

"Create Terraform for Gemini endpoint deployment"
"Deploy vector search for RAG application"
"Set up Vertex AI Pipeline for model training"
"Create Feature Store for ML features"
"Deploy custom model to Vertex AI endpoint"

Terraform Module Structure

1. Gemini API Endpoint

# gemini_endpoint.tf

# Gemini 2.0 Flash endpoint
resource "google_vertex_ai_endpoint" "gemini_endpoint" {
  display_name = "gemini-2-0-flash-endpoint"
  location     = var.region
  project      = var.project_id

  description = "Production Gemini 2.0 Flash endpoint"

  # Network configuration
  network = google_compute_network.vertex_vpc.id

  # Encryption
  encryption_spec {
    kms_key_name = google_kms_crypto_key.model_key.id
  }
}
…

2. Vector Search Infrastructure

# vector_search.tf

# Vector index for embeddings
resource "google_vertex_ai_index" "embeddings_index" {
  display_name = "${var.app_name}-embeddings-index"
  region       = var.region
  project      = var.project_id

  description = "Vector search index for RAG application"

  metadata {
    contents_delta_uri = google_storage_bucket.embeddings.url

    config {
      dimensions                  = 768  # text-embedding-gecko dimensions
      approximate_neighbors_count = 150
      distance_measure_type       = "DOT_PRODUCT_DISTANCE"
…