Google Cloud Python (google-cloud-storage / google-cloud-aiplatform)

Warnings

• gotcha There is no single 'google-cloud-python' package on PyPI. It must be installed per-service. 'pip install google-cloud-python' installs nothing useful — it's a meta-package stub that does not bring in any service clients.
  Fix: pip install google-cloud-storage, google-cloud-bigquery, google-cloud-aiplatform, etc. individually.
• gotcha DefaultCredentialsError is raised when no ADC credentials are found. Common in CI/CD when GOOGLE_APPLICATION_CREDENTIALS is not set and gcloud is not authenticated. On GCP services (Cloud Run, GKE, etc.) the metadata server provides credentials automatically.
  Fix: Set GOOGLE_APPLICATION_CREDENTIALS=/path/to/key.json, or run 'gcloud auth application-default login' locally.
• gotcha Service account JSON key files contain private keys — committing them to source control is a critical security issue. Google's Secret Manager or Workload Identity Federation are the recommended alternatives for production.
  Fix: Use Workload Identity Federation on GCP. For other clouds, use Secret Manager or environment-injected credentials.
• gotcha Vertex AI (google-cloud-aiplatform) requires vertexai.init(project=..., location=...) before any model calls. Missing location= silently defaults to us-central1 — models not available in that region return 404 or permission errors.
  Fix: Always call vertexai.init(project='PROJECT_ID', location='REGION') with an explicit region before using GenerativeModel.
• gotcha google-cloud-aiplatform ships weekly. The Vertex AI generative models API (GenerativeModel, generate_content) was added around v1.38 — older pinned versions don't have it and raise ImportError.
  Fix: pip install google-cloud-aiplatform>=1.38.0 for Gemini/GenerativeModel access.
• gotcha google-auth, google-api-core, and protobuf version conflicts are extremely common when multiple google-cloud packages are installed. pip may silently downgrade shared dependencies, breaking other packages.
  Fix: Install all google-cloud packages together in one pip command to let the resolver find compatible versions. Use pip check after installation.

Install

• pip install google-cloud-storage Cloud Storage
• pip install google-cloud-bigquery BigQuery
• pip install google-cloud-aiplatform Vertex AI / Gemini

Imports

• google.cloud.storage (ADC)

  from google.cloud import storage
  
  # ADC: reads GOOGLE_APPLICATION_CREDENTIALS env var or gcloud auth
  client = storage.Client()
  
  # Explicit service account
  from google.oauth2 import service_account
  credentials = service_account.Credentials.from_service_account_file('key.json')
  client = storage.Client(credentials=credentials, project='my-project')

  Direct module import path works but is unconventional. Always use 'from google.cloud import storage'.
• google.cloud.aiplatform (Vertex AI)

  import vertexai
  from vertexai.generative_models import GenerativeModel
  
  vertexai.init(project='my-project', location='us-central1')
  model = GenerativeModel('gemini-2.0-flash')
  response = model.generate_content('Hello')

  Vertex AI requires both project AND location at init. Missing location defaults to us-central1 but many models are only available in specific regions.

Quickstart

ADC credential chain: (1) GOOGLE_APPLICATION_CREDENTIALS JSON file, (2) gcloud application-default credentials, (3) GCE/Cloud Run metadata server.

import os
from google.cloud import storage

# Auth via GOOGLE_APPLICATION_CREDENTIALS env var pointing to service account JSON
# Or run 'gcloud auth application-default login' locally

client = storage.Client(project=os.environ['GOOGLE_CLOUD_PROJECT'])

# List buckets
buckets = list(client.list_buckets())
for bucket in buckets:
    print(bucket.name)

# Upload a file
bucket = client.bucket('my-bucket')
blob = bucket.blob('hello.txt')
blob.upload_from_string('Hello, GCS!')
print(f'Uploaded to gs://my-bucket/hello.txt')