LastLLM

LastLLM is a unified client for interacting with various LLM (Large Language Model) providers. It provides a consistent interface for text generation, structured data generation, and tool calling across different LLM services.

Features

• Unified interface for multiple LLM providers:
  • OpenAI
  • Anthropic
  • Google Gemini
  • Deepseek
  • Ollama
• Text generation
• Structured data generation with schema validation
• Tool/function calling support
• Rails integration
• Configuration management
• Error handling

Installation

Add this line to your application's Gemfile:

gem 'last_llm'

Configuration

Configure LastLLM globally or per instance:

LastLLM.configure do |config|
  config.default_provider = :openai
  config.default_model = 'gpt-3.5-turbo'

  # Configure OpenAI
  config.configure_provider(:openai, api_key: 'your-api-key')

  # Configure Anthropic
  config.configure_provider(:anthropic, api_key: 'your-api-key')

  # Configure Ollama
  config.configure_provider(:ollama, host: 'http://localhost:11434')
end

Logging

LastLLM supports custom logging for monitoring requests, responses, and errors. You can configure a custom logger to integrate with your existing logging system:

LastLLM.configure do |config|
  # Use Rails logger
  config.logger = Rails.logger

  # Set log level (optional, defaults to :info)
  config.log_level = :debug
end

In a Rails application, you can also configure this in your config/last_llm.yml:

logger: Rails.logger
log_level: debug

This will log LLM requests, responses, and errors to your Rails application logs, helping you debug and monitor your AI interactions.

Rails Integration

LastLLM automatically integrates with Rails applications. Create a config/last_llm.yml file:

default_provider: openai
default_model: gpt-3.5-turbo
globals:
  timeout: 30
  max_retries: 3

Usage

Basic Text Generation

client = LastLLM.client
response = client.generate_text("What is the capital of France?")

Structured Data Generation

schema = Dry::Schema.JSON do
  required(:name).filled(:string)
  required(:age).filled(:integer)
  optional(:email).maybe(:string)
end

client = LastLLM.client
result = client.generate_object("Generate information about a person", schema)

Tool Calling

calculator = LastLLM::Tool.new(
  name: "calculator",
  description: "Perform mathematical calculations",
  parameters: {
    type: "object",
    properties: {
      operation: {
        type: "string",
        enum: ["add", "subtract", "multiply", "divide"]
      },
      a: { type: "number" },
      b: { type: "number" }
    },
    required: ["operation", "a", "b"]
  },
  function: ->(params) {
    case params[:operation]
    when "add"
      { result: params[:a] + params[:b] }
    end
  }
)

Error Handling

LastLLM provides several error classes:

• LastLLM::Error: Base error class
• LastLLM::ConfigurationError: Configuration-related errors
• LastLLM::ValidationError: Schema validation errors
• LastLLM::ApiError: API request errors

Development

After checking out the repo, run bundle install to install dependencies. Then, run rspec to run the tests.

Contributing

Bug reports and pull requests are welcome on GitHub.