Google (Gemini & Vertex AI)

Gemini is available two ways: through Google AI (google, an API key) or through Vertex AI / Agent Platform (google-vertex, Application Default Credentials). Pick whichever matches how your org buys Google models.

Google AI (Gemini API key)

The quickest path: a single API key.

Create a key in Google AI Studio.
Store it as GEMINI_API_KEY.

model: google/gemini-2.5-pro
providers:
  allowlist: [google]

- uses: louisescher/crabd@v0
  with:
    model: google/gemini-2.5-pro
  env:
    GEMINI_API_KEY: ${{ secrets.GEMINI_API_KEY }}

Vertex AI / Agent Platform

Use google-vertex when you want Gemini billed and governed through Google Cloud. It authenticates with Application Default Credentials, a service-account key the runner can read.

Create a service account with Vertex AI access and download its JSON key (or use google-github-actions/auth to provide ADC).
Provide the ADC file path and your project/location via env.

model: google-vertex/gemini-2.5-pro
providers:
  allowlist: [google-vertex]

- uses: google-github-actions/auth@v2
  with:
    credentials_json: ${{ secrets.GCP_SA_KEY }}
- uses: louisescher/crabd@v0
  with:
    model: google-vertex/gemini-2.5-pro
  env:
    GOOGLE_APPLICATION_CREDENTIALS: ${{ steps.auth.outputs.credentials_file_path }}
    GOOGLE_CLOUD_PROJECT: my-gcp-project
    GOOGLE_CLOUD_LOCATION: us-central1