Skip to content
Codemetry

Google AI

Google’s Gemini models offer strong reasoning capabilities with good integration into Google Cloud environments.

Setup

Section titled “Setup”

  1. Get an API Key

    Visit aistudio.google.com/apikey to create an API key.

    Alternatively, use Google Cloud’s Vertex AI for enterprise deployments.

  2. Set Environment Variable

    Terminal window
    export GOOGLE_AI_API_KEY=AIza...

    Or add to your .env file:

    GOOGLE_AI_API_KEY=AIza...

  3. Configure Codemetry

    config/codemetry.php
    'ai' => [
        'enabled' => true,
        'engine' => 'google',
        'model' => null,  // Uses gemini-3.8-flash by default
    ],

  4. Run Analysis

    Terminal window
    php artisan codemetry:analyze --days=7 --ai=1 --ai-engine=google

Available Models

Section titled “Available Models”
Model Default Cost Quality Speed
gemini-3.8-flash ✓ Low Good Very Fast
gemini-3.5-flash-lite Very Low Good Very Fast
gemini-4-argon High Best Medium

Google retires models regularly. If a model on this page stops working, check the provider’s model list and set a current one.

Choosing a Model

Section titled “Choosing a Model”

For routine analysis: Use gemini-3.8-flash (default)

  • Google’s current Flash model: fast and inexpensive

Alternative: Use gemini-3.5-flash-lite

  • The lowest-cost option for frequent runs

For critical analyses: Use gemini-4-argon

  • Google’s Pro-tier model for deeper investigations

There is no --ai-model flag. To override the model for a single run, set CODEMETRY_AI_MODEL:

Terminal window
CODEMETRY_AI_MODEL=gemini-3.5-flash-lite php artisan codemetry:analyze --days=7 --ai=1 --ai-engine=google

Or in configuration:

'ai' => [
    'engine' => 'google',
    'model' => 'gemini-3.5-flash-lite',
],

API Key Options

Section titled “API Key Options”

Codemetry checks for API keys in this order:

  1. CODEMETRY_AI_API_KEY — Unified key for any engine
  2. GOOGLE_AI_API_KEY — Google AI-specific key
Terminal window
# Option 1: Unified
export CODEMETRY_AI_API_KEY=AIza...


# Option 2: Engine-specific
export GOOGLE_AI_API_KEY=AIza...

Cost Management

Section titled “Cost Management”

Google charges per token. Codemetry sends minimal data:

  • ~500-1000 input tokens per day analyzed
  • ~200-500 output tokens per summary

Prices change often, so check Gemini API pricing for current per-token rates.

Google Cloud Integration

Section titled “Google Cloud Integration”

For enterprise deployments, consider using Vertex AI:

  1. Enable Vertex AI API in Google Cloud Console
  2. Set up authentication (service account or ADC)
  3. Use Vertex AI endpoints instead of AI Studio

This provides:

  • Enterprise SLAs
  • VPC integration
  • Compliance certifications
  • Centralized billing

Troubleshooting

Section titled “Troubleshooting”

“Invalid API key”

Section titled ““Invalid API key””

Cause: API key is incorrect or not enabled.

Solution:

“Quota exceeded”

Section titled ““Quota exceeded””

Cause: Free tier limits reached.

Solution:

  • Wait for quota reset (resets periodically)
  • Enable billing for higher limits
  • Use Flash models for lower quota consumption

“Model not found”

Section titled ““Model not found””

Cause: Model name incorrect or model deprecated.

Solution:

  • Check current model names in Google AI documentation
  • Model names may change as versions update
  • Use a different model variant

“Region not supported”

Section titled ““Region not supported””

Cause: Google AI not available in your region.

Solution:

  • Check Google AI availability for your region
  • Consider using a VPN for development
  • Use a different AI engine

Slow Responses with Pro Model

Section titled “Slow Responses with Pro Model”

Symptom: gemini-4-argon takes longer than expected.

Solution:

  • This is normal; Pro models prioritize quality over speed
  • Use gemini-3.8-flash (the default) for faster responses
  • Consider Pro only for detailed investigations