Google AI Setup¶

Configuration guide for Gemini API and Vertex AI integration.

Prerequisites¶

For Gemini API¶

Google AI Studio account
API key from aistudio.google.com

For Vertex AI¶

Google Cloud Platform account
Project with Vertex AI API enabled
Service account with appropriate permissions (or API key)

Environment Variables¶

# Gemini API (Backend)
GEMINI_API_KEY="your-gemini-api-key"

# Vertex AI (Compute Server)
GOOGLE_API_KEY="your-google-api-key"
GOOGLE_CLOUD_PROJECT="your-gcp-project-id"
GOOGLE_CLOUD_LOCATION="global"  # or specific region
GOOGLE_APPLICATION_CREDENTIALS="/path/to/service-account.json"
GOOGLE_GENAI_USE_VERTEXAI="true"

# Model Configuration
VERTEX_IMAGE_MODEL="gemini-3-pro-image-preview"
VERTEX_TEXT_MODEL="gemini-3-pro-preview"

Backend Configuration (Gemini API)¶

# core/app/core/config.py
class Settings(BaseSettings):
    GEMINI_API_KEY: str | None = None

    # Model defaults
    CAPTIONING_AI_PROVIDER: str = "gemini"
    CAPTIONING_MODEL: str = "gemini-2.5-flash-preview-05-20"

    STYLING_AI_PROVIDER: str = "gemini"
    STYLING_MODEL: str = "gemini-2.0-flash-latest"

    LLM_VISIBILITY_AI_PROVIDER: str = "gemini"
    LLM_VISIBILITY_MODEL: str = "gemini-flash-latest"

Compute Server Configuration (Vertex AI)¶

# app/core/config.py
class Settings(BaseSettings):
    # Authentication
    GOOGLE_API_KEY: Optional[str] = None
    GOOGLE_CLOUD_PROJECT: Optional[str] = None
    GOOGLE_CLOUD_LOCATION: str = "global"
    GOOGLE_APPLICATION_CREDENTIALS: Optional[str] = None
    GOOGLE_GENAI_USE_VERTEXAI: bool = True

    # Models
    VERTEX_IMAGE_MODEL: str = "gemini-3-pro-image-preview"
    VERTEX_TEXT_MODEL: str = "gemini-3-pro-preview"

Dependencies¶

Backend (Gemini API)¶

# pyproject.toml
[project.dependencies]
google-generativeai = ">=0.8.5"
google-auth = ">=2.40.3"
litellm = "^1.75.8"

Compute Server (Vertex AI)¶

# pyproject.toml
[project.dependencies]
google-genai = "^1.0.0"
google-auth = ">=2.40.3"

Client Initialization¶

Gemini API (via GeminiService)¶

import google.generativeai as genai
from google.genai import Client, types

class GeminiService:
    def __init__(self, api_key: str, model: str):
        self.api_key = api_key
        self.model = model
        self.client = Client(api_key=api_key)

    def generate_content(self, contents, system_instruction=None):
        config = types.GenerateContentConfig(
            response_mime_type="application/json",
            system_instruction=[types.Part.from_text(system_instruction)] if system_instruction else None
        )
        return self.client.models.generate_content(
            model=self.model,
            contents=contents,
            config=config,
        )

Vertex AI (via VertexProcessor)¶

from google import genai
from google.genai import types

class VertexProcessor:
    def _create_client(self) -> genai.Client:
        # Priority 1: Service Account (IAM)
        if settings.GOOGLE_APPLICATION_CREDENTIALS:
            return genai.Client(vertexai=True)

        # Priority 2: Vertex AI with API Key + Project
        if settings.GOOGLE_GENAI_USE_VERTEXAI:
            return genai.Client(
                api_key=settings.GOOGLE_API_KEY,
                vertexai=True,
                project=settings.GOOGLE_CLOUD_PROJECT,
                location=settings.GOOGLE_CLOUD_LOCATION,
            )

        # Fallback: Standard Gemini API
        return genai.Client(api_key=settings.GOOGLE_API_KEY)

Authentication Priority¶

Vertex AI Authentication Order¶

Service Account (IAM) - GOOGLE_APPLICATION_CREDENTIALS set
Most secure for production
Uses GCP IAM roles
API Key + Project - GOOGLE_GENAI_USE_VERTEXAI=true
Requires GOOGLE_API_KEY and GOOGLE_CLOUD_PROJECT
Good for development
Standard Gemini API - Fallback
Uses GOOGLE_API_KEY only
No image generation support

Model Aliases (Gemini API)¶

model_aliases = {
    "gemini": "gemini-2.0-flash",
    "gemini-2.5-flash": "gemini-2.5-flash-preview-05-20",
    "gemini-2.0-flash": "gemini-2.0-flash-latest",
    "gemini-1.5-pro": "gemini-1.5-pro-latest",
    "gemini-1.5-flash": "gemini-1.5-flash-latest",
    "gemini-flash-latest": "gemini-flash-latest",
}

Default Parameters¶

Gemini API Defaults¶

Parameter	Value	Description
`max_tokens`	8192	Maximum output
`temperature`	0.4	Balanced creativity
`top_p`	0.95	Nucleus sampling
`top_k`	40	Gemini-specific

Vertex AI Defaults¶

Parameter	Value	Description
`temperature`	0.7	Generation creativity
`top_p`	0.95	Nucleus sampling
`top_k`	40	Token selection

Verification¶

Test Gemini API¶

from google.genai import Client

client = Client(api_key=settings.GEMINI_API_KEY)
response = client.models.generate_content(
    model="gemini-2.0-flash",
    contents="Hello, Gemini!"
)
print(response.text)

Test Vertex AI¶

from google import genai

client = genai.Client(
    api_key=settings.GOOGLE_API_KEY,
    vertexai=True,
    project=settings.GOOGLE_CLOUD_PROJECT,
    location=settings.GOOGLE_CLOUD_LOCATION,
)

response = client.models.generate_content(
    model="gemini-3-pro-preview",
    contents="Hello from Vertex AI!"
)
print(response.text)

Troubleshooting¶

Invalid API Key¶

Error: API key not valid

Verify key is correct and active
Check key has access to required models

Quota Exceeded¶

Error: Quota exceeded for quota metric

Check usage in Google Cloud Console
Upgrade quota or implement rate limiting

Safety Filter Blocked¶

Error: Response blocked by safety filters

For Vertex AI: Safety settings are disabled by default
For Gemini API: Rephrase prompt or adjust thresholds

Model Not Found¶

Error: Model not found

Verify model name is correct
Check model availability in your region
Ensure Vertex AI API is enabled in project