Google Vertex AI Setup

Overview

Google Cloud Vertex AI provides access to state-of-the-art models (Gemini, PaLM) with enterprise-grade features.

Prerequisites

1. Google Cloud Project with billing enabled
2. gcloud CLI installed: https://cloud.google.com/sdk/docs/install
3. Python 3.8+

Step 1: Create a Google Cloud Project

# Create project
gcloud projects create my-llm-project --name="LLM Parameter Testing"

# Set default project
gcloud config set project my-llm-project

# Enable Vertex AI API
gcloud services enable aiplatform.googleapis.com

Step 2: Authenticate

# Login and set up Application Default Credentials
gcloud auth application-default login

This creates credentials that the Python SDK will use automatically.

Step 3: Configure Environment

Edit your .env file:

# Google Cloud
GOOGLE_PROJECT_ID=my-llm-project
GOOGLE_REGION=us-central1
VERTEX_MODEL=gemini-1.5-pro  # or gemini-pro, text-bison, etc.

Step 4: Supported Models

As of May 2026, Vertex AI supports:

• Gemini 1.5 Pro – Latest multimodal model
• Gemini 1.5 Flash – Fast, efficient variant
• Gemini 2.0 (preview) – Next-generation
• Claude 3 (via Bedrock integration)

List available models:

gcloud ai models list --region=us-central1

Step 5: Run an Experiment

# Activate venv and install deps
source .venv-exp/bin/activate
pip install -r requirements-experiments.txt

# Run preset sweep
python3 scripts/cloud/vertex_sweep.py --preset summarization --trials 3

# Analyze results
python3 scripts/analyze_results.py --input runs/latest.jsonl

Cost Estimates

Vertex AI pricing varies by region and model. Typical costs for 1000 API calls:

• Gemini 1.5 Flash: ~$0.075
• Gemini 1.5 Pro: ~$0.30
• Text-Bison: ~$0.01

Current pricing

Troubleshooting

"Permission denied" error:

gcloud projects add-iam-policy-binding my-llm-project \
  --member=user:your-email@gmail.com \
  --role=roles/aiplatform.user

"Model not found" error: Ensure the model is available in your region:

gcloud ai models list --region=us-central1 | grep gemini

"Quota exceeded" error: Check and increase quotas:

gcloud compute project-info describe --project=my-llm-project \
  --format='value(quotas)'

---

Next: Run your first experiment