A lightweight Ruby framework for orchestrating LLM operations with OpenAI and Anthropic Claude. This gem provides a simple way to:
- Manage prompt templates
- Chain LLM operations
- Maintain conversation context
- Switch between OpenAI and Claude providers
Add this line to your application's Gemfile:
gem 'llm_orchestrator'And then execute:
$ bundle installConfigure your API keys:
LlmOrchestrator.configure do |config|
config.openai_api_key = ENV['OPENAI_API_KEY']
config.claude_api_key = ENV['CLAUDE_API_KEY']
config.default_llm_provider = :claude # or :openai
end# Create and use a prompt template
prompt = LlmOrchestrator::Prompt.new("Answer this {type} question: {question}")
formatted = prompt.format(
type: "math",
question: "What is 2+2?"
)# Using OpenAI
openai = LlmOrchestrator::OpenAI.new
response = openai.generate("What is 2+2?", model: "gpt-3.5-turbo")
# Using Claude
claude = LlmOrchestrator::Anthropic.new
response = claude.generate("What is 2+2?", model: "claude-3-opus-20240229")# Create a chain with memory
memory = LlmOrchestrator::Memory.new
chain = LlmOrchestrator::Chain.new(memory: memory)
# Add processing steps
chain.add_step do |input, memory|
llm = LlmOrchestrator::Anthropic.new
llm.generate(input, context: memory.context_string)
end
# Run multiple interactions with context
result1 = chain.run("What is the capital of France?")
result2 = chain.run("What is its population?") # Uses previous context
# Clear memory when needed
chain.clear_memoryAfter checking out the repo, run:
$ bundle install
$ bundle exec rspecThe tests use VCR to record HTTP interactions. To run the tests with real API calls:
- Set up your environment variables:
export OPENAI_API_KEY="your-key-here"
export CLAUDE_API_KEY="your-key-here"- Delete the VCR cassettes (if they exist):
rm -rf spec/fixtures/vcr_cassettes- Run the tests:
bundle exec rspecBug reports and pull requests are welcome on GitHub at https://github.com/[USERNAME]/llm_orchestrator.
The gem is available as open source under the terms of the MIT License.