lex-gemini

Google Gemini API integration for LegionIO. Generate content, create embeddings, manage files, count tokens, and cache content using Google's Gemini models.

gem install lex-gemini

Or add to your Gemfile:

gem 'lex-gemini'

• generate - Generate text content using Gemini models
• stream_generate - Stream generated content

• embed - Generate embedding vectors from text
• batch_embed - Generate multiple embeddings in one request

• list - List available Gemini models
• get - Get details about a specific model

• count - Count tokens in content

• upload - Upload a file for use with Gemini
• list - List uploaded files
• get - Get file metadata
• delete - Delete an uploaded file

• create - Create cached content for repeated use
• list - List cached contents
• get - Get cached content details
• update - Update cached content expiration
• delete - Delete cached content

Set your API key in your LegionIO settings:

{
  "gemini": {
    "api_key": "AIza..."
  }
}

require 'legion/extensions/gemini/helpers/client'

client = Legion::Extensions::Gemini::Helpers::Client.new(
  api_key: ENV['GEMINI_API_KEY'],
  model: 'gemini-2.0-flash'
)

# Generate content
result = client.generate_content(
  contents: [{ parts: [{ text: 'Explain quantum entanglement in one sentence.' }] }]
)
puts result.dig('candidates', 0, 'content', 'parts', 0, 'text')

# Generate embeddings
embedding_client = Legion::Extensions::Gemini::Helpers::Client.new(
  api_key: ENV['GEMINI_API_KEY'],
  model: 'gemini-embedding-exp'
)
embedding = embedding_client.embed_content(
  content: { parts: [{ text: 'Hello world' }] }
)
puts embedding['embedding']['values'].length

# Count tokens
tokens = client.count_tokens(
  contents: [{ parts: [{ text: 'How many tokens?' }] }]
)
puts tokens['totalTokens']

• faraday >= 2.0 - HTTP client
• faraday-multipart (optional) - required only for multipart file uploads; raw binary upload used as fallback

• Ruby >= 3.4
• LegionIO framework (optional for standalone usage)
• Google Gemini API key (Get one here)

MIT