Aigen::Google

Ruby SDK for Google's Gemini API - build AI-powered applications with text generation, chat, streaming, and multimodal content support.

Features

• Text Generation: Generate content from text prompts
• Chat with History: Multi-turn conversations with context preservation
• Streaming: Real-time response delivery with progressive chunks
• Multimodal Content: Combine text and images in requests
• Generation Config: Control output with temperature, top_p, top_k, max_output_tokens
• Safety Settings: Configure content filtering for harm categories
• Comprehensive Error Handling: Automatic retries with exponential backoff for rate limits and server errors
• Ruby 3.1+ Support: Modern Ruby idioms with frozen string literals

Installation

Add this line to your application's Gemfile:

gem 'aigen-google'

And then execute:

bundle install

Or install it yourself as:

gem install aigen-google

Quick Start

require 'aigen/google'

# Configure with your API key
Aigen::Google.configure do |config|
  config.api_key = ENV['GOOGLE_API_KEY']
end

# Generate content
client = Aigen::Google::Client.new
response = client.generate_content(prompt: "Explain quantum computing in simple terms")
puts response["candidates"][0]["content"]["parts"][0]["text"]

Configuration

Block-based Configuration (Recommended)

Aigen::Google.configure do |config|
  config.api_key = ENV['GOOGLE_API_KEY']
  config.default_model = "gemini-pro"
  config.timeout = 60
end

client = Aigen::Google::Client.new

Instance-based Configuration

client = Aigen::Google::Client.new(
  api_key: "your-api-key",
  model: "gemini-pro",
  timeout: 60
)

Available Configuration Options

Option	Type	Default	Description
api_key	String	ENV['GOOGLE_API_KEY']	Your Google AI API key (Get one here)
default_model	String	"gemini-pro"	Default model to use for requests
timeout	Integer	30	HTTP request timeout in seconds
retry_count	Integer	3	Number of retry attempts for rate limits/server errors

Usage Examples

Basic Text Generation

client = Aigen::Google::Client.new

response = client.generate_content(prompt: "Write a haiku about Ruby programming")
text = response["candidates"][0]["content"]["parts"][0]["text"]
puts text

Chat with History

chat = client.start_chat

# First message
chat.send_message("What is Ruby?")

# Follow-up with context
response = chat.send_message("What are its main features?")
text = response["candidates"][0]["content"]["parts"][0]["text"]
puts text

# View conversation history
puts chat.history

Streaming Responses

Stream responses as they're generated for real-time feedback:

# With block (immediate processing)
client.generate_content_stream(prompt: "Tell me a story") do |chunk|
  text = chunk["candidates"][0]["content"]["parts"][0]["text"]
  print text
end

# With Enumerator (lazy evaluation)
stream = client.generate_content_stream(prompt: "Count to 10")
stream.each do |chunk|
  text = chunk["candidates"][0]["content"]["parts"][0]["text"]
  print text
end

Chat Streaming

chat = client.start_chat

chat.send_message_stream("Tell me a joke") do |chunk|
  text = chunk["candidates"][0]["content"]["parts"][0]["text"]
  print text
end
# History is updated after streaming completes

Multimodal Content (Text + Images)

require 'base64'

# Prepare image data
image_data = Base64.strict_encode64(File.read("photo.jpg"))

# Create multimodal content
text_content = Aigen::Google::Content.text("What is in this image?")
image_content = Aigen::Google::Content.image(
  data: image_data,
  mime_type: "image/jpeg"
)

# Generate content with both
response = client.generate_content(contents: [text_content.to_h, image_content.to_h])
text = response["candidates"][0]["content"]["parts"][0]["text"]
puts text

Configuring Generation Parameters

Control output quality and randomness:

response = client.generate_content(
  prompt: "Write a creative story",
  temperature: 0.9,      # Higher = more creative (0.0-1.0)
  top_p: 0.95,           # Nucleus sampling (0.0-1.0)
  top_k: 40,             # Top-k sampling
  max_output_tokens: 1024 # Maximum response length
)

Safety Settings

Configure content filtering:

# Use default settings (BLOCK_MEDIUM_AND_ABOVE for all categories)
settings = Aigen::Google::SafetySettings.default

# Or customize
settings = Aigen::Google::SafetySettings.new([
  {
    category: Aigen::Google::SafetySettings::HARM_CATEGORY_HATE_SPEECH,
    threshold: Aigen::Google::SafetySettings::BLOCK_LOW_AND_ABOVE
  },
  {
    category: Aigen::Google::SafetySettings::HARM_CATEGORY_DANGEROUS_CONTENT,
    threshold: Aigen::Google::SafetySettings::BLOCK_MEDIUM_AND_ABOVE
  }
])

response = client.generate_content(
  prompt: "Hello",
  safety_settings: settings.to_h
)

Image Generation (Nano Banana 🍌)

Generate images using Gemini's image generation models (nicknamed "Nano Banana"):

# Use an image generation model
client = Aigen::Google::Client.new(model: 'gemini-2.5-flash-image')

# Simple image generation
response = client.generate_image("A serene mountain landscape at sunset")

if response.success?
  response.save("landscape.png")
  puts "Image saved!"
  puts "Description: #{response.text}"
else
  puts "Generation failed: #{response.failure_message}"
end

With Size and Aspect Ratio

response = client.generate_image(
  "A futuristic cityscape with flying cars",
  aspect_ratio: "16:9",  # Options: "1:1", "16:9", "9:16", "4:3", "3:4", "5:4", "4:5"
  size: "2K"             # Options: "1K", "2K", "4K"
)

response.save("city.png") if response.success?

ImageResponse Helper Methods

The generate_image method returns an ImageResponse object with convenient helpers:

response = client.generate_image("A cute puppy")

# Check success
response.success?              # => true/false
response.has_image?            # => true/false

# Access image data
response.image_data            # => Binary image data
response.mime_type             # => "image/png"
response.text                  # => Text description

# Save image
response.save("output.png")

# Handle failures
unless response.success?
  puts response.failure_reason   # => "IMAGE_OTHER", "SAFETY", etc.
  puts response.failure_message  # => Detailed error message
end

# Access raw response
response.raw_response           # => Full API response hash

Available Models

• gemini-2.5-flash-image - Standard image generation (up to 2K)
• gemini-2.0-flash-exp - Experimental with native image generation
• gemini-3-pro-image-preview - Preview with 4K support

Error Handling

The SDK provides comprehensive error handling with automatic retries:

Exception Types

Exception	When Raised	Retry Behavior
Aigen::Google::AuthenticationError	Invalid API key (401/403)	No retry
Aigen::Google::InvalidRequestError	Malformed request (400/404)	No retry
Aigen::Google::RateLimitError	Rate limit exceeded (429)	Automatic retry with backoff
Aigen::Google::ServerError	Server error (500-599)	Automatic retry with backoff
Aigen::Google::TimeoutError	Request timeout	Automatic retry with backoff
Aigen::Google::ApiError	Other API errors	No retry

Retry Behavior

The SDK automatically retries rate limit (429) and server errors (500-599) with exponential backoff:

• Attempt 1: Wait 1 second, retry
• Attempt 2: Wait 2 seconds, retry
• Attempt 3: Wait 4 seconds, raise error if fails

Maximum retry attempts: 3 (configurable via retry_count)

Handling Errors

begin
  response = client.generate_content(prompt: "Hello")
rescue Aigen::Google::AuthenticationError => e
  puts "Invalid API key: #{e.message}"
  puts "Get your API key at: https://makersuite.google.com/app/apikey"
rescue Aigen::Google::RateLimitError => e
  puts "Rate limit exceeded: #{e.message}"
  puts "Try reducing request rate or increasing quota"
rescue Aigen::Google::ServerError => e
  puts "Server error (#{e.status_code}): #{e.message}"
  puts "Try again later - the API may be experiencing issues"
rescue Aigen::Google::TimeoutError => e
  puts "Request timed out: #{e.message}"
rescue Aigen::Google::ApiError => e
  puts "API error (#{e.status_code}): #{e.message}"
end

API Reference

Full API documentation is available at https://rubydoc.info/gems/aigen-google (YARD documentation).

Key Classes

• Aigen::Google::Client - Main client for API requests
• Aigen::Google::Chat - Multi-turn conversation management
• Aigen::Google::Content - Multimodal content builder
• Aigen::Google::GenerationConfig - Generation parameter configuration
• Aigen::Google::SafetySettings - Content filtering configuration
• Aigen::Google::Configuration - Global configuration

Development

After checking out the repo, run bin/setup to install dependencies. Then, run rake spec to run the tests. You can also run bin/console for an interactive prompt that will allow you to experiment.

Running Tests

# Run all tests
bundle exec rspec

# Run with coverage report
COVERAGE=true bundle exec rspec

# Run linter
bundle exec rake standard

# Run both tests and linter
bundle exec rake

Code Quality

• Test Coverage: >= 90% (enforced with SimpleCov)
• Code Style: StandardRB (0 offenses required)
• Ruby Version: >= 3.1.0

Contributing

Bug reports and pull requests are welcome on GitHub at https://github.com/neuralmux/aigen-google.

1. Fork the repository
2. Create your feature branch (git checkout -b my-new-feature)
3. Commit your changes (git commit -am 'Add some feature')
4. Push to the branch (git push origin my-new-feature)
5. Create a new Pull Request

License

The gem is available as open source under the terms of the MIT License.

Changelog

See CHANGELOG.md for version history and release notes.

Resources

• Gemini API Documentation
• Get your API Key
• Ruby Style Guide
• YARD Documentation Guide