Welcome to Getting Started with RubyLLM

RubyLLM is a Ruby gem that gives you a single, elegant interface to work with multiple AI providers — OpenAI, Anthropic, Google Gemini, and more. Instead of learning a different SDK for each provider, you write Ruby. One API, any model.

This tutorial runs entirely in your browser — no installation, no API keys required to follow along. The Ruby runtime is built into the page.

What You’ll Learn

By the end of this tutorial, you’ll be able to:

• Configure RubyLLM and connect it to multiple AI providers
• Explore the model registry to discover and compare available models
• Build multi-turn conversations with persistent context
• Stream responses in real time as the model generates them
• Send files and images alongside messages with multi-modal input
• Generate embeddings and work with vector representations of text
• Generate images from text prompts with DALL-E and Imagen
• Transcribe audio to text with Whisper and GPT-4o Transcribe
• Create AI tools using plain Ruby classes — no JSON schema boilerplate
• Build agents that reason, plan, and call tools autonomously
• Extract structured data from unstructured text using typed outputs
• Compose agentic workflows — sequential pipelines and evaluation loops

Prerequisites

This tutorial assumes you’re comfortable writing Ruby. You don’t need any prior experience with AI APIs or LLMs — that’s what we’re here for.

If you’d like a Ruby refresher, The Odin Project and ruby.dev are great starting points.

What Makes RubyLLM Different?

Most AI SDKs expose you to the raw HTTP API: you build message arrays, handle token limits, parse JSON, and manage state yourself. RubyLLM abstracts all of that:

• One interface, many providers — switch from GPT-4o to Claude Sonnet by changing a model name string
• Conversations are objects — history, context, and state are managed for you
• Tools are Ruby classes — define a tool the same way you’d write any Ruby class
• Streaming is a block — chat.ask("...") { |chunk| print chunk.text }

Let’s get started.