Gemini CLI instructions for Ubuntu Nobel 24.04.3

01-gemini.sh

# Install Gemini CLI https://github.com/google-gemini/gemini-cli
apt-cache policy nodejs
# if distro is using node < 20 then manage versions using nvm instead https://github.com/nvm-sh/nvm/tree/master
curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.40.3/install.sh | bash
# open new shell or copy/paste export/load commands
nvm install 22  # current LTS from https://nodejs.org/en/about/previous-releases
npm install -g @google/gemini-cli

# API keys not working in gemini-cli >= 0.1.15 https://aistudio.google.com/u/0/apikey
#export GEMINI_API_KEY="AIzaSyCRSFCdFFn8G_REDACTED-..." # gregbray-vpc

# instead use GCP api key from GSA https://cloud.google.com/vertex-ai/generative-ai/docs/start/api-keys
# Note: this same option should work for Claude Code https://docs.anthropic.com/en/docs/claude-code/google-vertex-ai
gcloud iam service-accounts create gemini-cli --project=gregbray-vpc \
  --description="Exported Service Account for gemini-cli"
gcloud projects add-iam-policy-binding gregbray-vpc \
  --member="serviceAccount:[email protected]" \
  --role=roles/aiplatform.expressUser --condition None
gcloud services api-keys create --display-name="Gemini CLI" --key-id=gemini-cli \
  --service-account=gemini-cli@gregbray-vpc.iam.gserviceaccount.com \
  --project gregbray-vpc

# Edit your .bashrc to look something like this (replacing the lines added by nvm):
if [[ -d "$HOME/.nvm" ]]; then
  export NVM_DIR="$HOME/.nvm"
  [ -s "$NVM_DIR/nvm.sh" ] && \. "$NVM_DIR/nvm.sh"  # This loads nvm
  [ -s "$NVM_DIR/bash_completion" ] && \. "$NVM_DIR/bash_completion"  # This loads nvm bash_completion

  # Configure gemini cli to use vertex api keys https://cloud.google.com/vertex-ai/generative-ai/docs/start/api-keys
  export GOOGLE_GENAI_USE_VERTEXAI=true
  export GOOGLE_CLOUD_PROJECT=gregbray-vpc
  export GOOGLE_CLOUD_LOCATION=us-central1
  export GOOGLE_API_KEY=AQ.Ab8RN6JM...REDACTED....
  
  # Settings for Crush to use Vertex AI via GCP ADC https://github.com/charmbracelet/crush
  export VERTEXAI_PROJECT=gregbray-vpc
  export VERTEXAI_LOCATION=global
  
  # Settings for Claude Code https://docs.anthropic.com/en/docs/claude-code/google-vertex-ai
  export CLAUDE_CODE_USE_VERTEX=1
  export CLOUD_ML_REGION=us-east5
  export ANTHROPIC_VERTEX_PROJECT_ID=gregbray-vpc

  #export ANTHROPIC_MODEL='claude-opus-4-1@20250805'
  #export ANTHROPIC_SMALL_FAST_MODEL='claude-3-5-haiku@20241022'
fi

# open gemini cli in a new project folder and select option 3 for vertex api
cd ~/code/new
gemini

02-crush.sh

# Crush is a golang based multi-model cli https://github.com/charmbracelet/crush/
# Install via npm or see git repo for other options (does not require nodejs/npm)
nvm use 22
npm install -g @charmland/crush

# make sure .bashrc includes these environment variables
export VERTEXAI_PROJECT=gregbray-vpc
export VERTEXAI_LOCATION=global

# Setup ADC (crush does not seem to use API keys, which is good)
# copy the url to a chrome session where you are logged in on argolis
gcloud auth application-default login --no-launch-browser

# Navigate to a project or new folder and start crush. Select Gemini 2.5 Pro or flash as the model
# Easiest to skip the project init stuff and go back to that later
cd ~/code/temp
crush

# after model is selected an also use
crush run "why is the sky blue"

# model selection is stored in ~/.local/share/crush/crush.json

# Add claude-sonnet model from https://cloud.google.com/vertex-ai/generative-ai/docs/partner-models/claude/sonnet-4
# Must enable model for your project using "View model card in Model Garden" first
mkdir -p ~/.config/crush
cat > ~/.config/crush/crush.json << EOF
{
  "$schema": "https://charm.land/crush.json",
  "lsp": {
    "Go": {
      "command": "gopls"
    }
  },
  "providers": {
    "vertexai": {
      "models": [
        {
          "id": "claude-sonnet-4@20250514",
          "name": "VertexAI: Sonnet 4",
          "cost_per_1m_in": 3,
          "cost_per_1m_out": 15,
          "cost_per_1m_in_cached": 3.75,
          "cost_per_1m_out_cached": 0.3,
          "context_window": 200000,
          "default_max_tokens": 50000,
          "can_reason": true,
          "supports_attachments": true
        }
      ]
    }
  }
}
EOF

03-claude-code.sh

# Claude Code. See full instructions at https://docs.anthropic.com/en/docs/claude-code/google-vertex-ai
# Install cli https://docs.anthropic.com/en/docs/claude-code/overview
nvm use 22
npm install -g @anthropic-ai/claude-code

# Navigate to your project
cd ~/code/temp

# Make sure the following models are enabled for your project using "view model card in Model Garden"
# https://cloud.google.com/vertex-ai/generative-ai/docs/partner-models/claude/sonnet-4
# https://cloud.google.com/vertex-ai/generative-ai/docs/partner-models/claude/haiku-3-5

# Make sure the following are set in .bashrc
export CLAUDE_CODE_USE_VERTEX=1
export CLOUD_ML_REGION=us-east5   # not sure why the new global endpoint isn't working
export ANTHROPIC_VERTEX_PROJECT_ID=gregbray-vpc

# Start Claude in a project folder
claude