You’ve Completed the RubyLLM Tutorial

You started with a blank Ruby environment and worked your way through the full surface area of RubyLLM — from a simple chat call to multi-step agentic workflows running entirely in your browser.

What You Built

Here’s everything you covered:

• Configuration — connected RubyLLM to AI providers and tuned model defaults
• Model Registry — explored and compared models across providers programmatically
• Chat & Conversations — built multi-turn conversations with persistent context
• Streaming — received responses token-by-token using a simple block interface
• Multimodal Input — sent images and files alongside text messages
• Embeddings — generated vector representations and measured semantic similarity
• Image Generation — created images from text prompts with DALL-E and Imagen
• Audio Transcription — converted speech to text with Whisper and GPT-4o Transcribe
• Tools — defined Ruby classes as callable tools with automatic schema generation
• Structured Output — extracted typed data from unstructured text
• Agents — built autonomous agents that reason, plan, and call tools in a loop
• Agentic Workflows — composed sequential pipelines and self-evaluating feedback loops

What to Build Next

RubyLLM is designed to drop into any Ruby context. Here are some natural next steps:

Add AI to a Rails app

def create
  chat = RubyLLM::Chat.new
  @response = chat.ask(params[:message])
  render json: { message: @response.content }
end

Build a RAG pipeline

Combine embeddings with a vector database (pgvector, Weaviate, Qdrant) to give your model access to your own documents.

Run tools in background jobs

Wrap long-running tool calls in Sidekiq or Solid Queue so agents can do real work without blocking your web process.

Keep in Touch

RubyLLM is open source and actively developed. If you hit a bug, have a feature idea, or want to contribute:

• GitHub: github.com/crmne/ruby_llm
• RubyGems: rubygems.org/gems/ruby_llm

Thanks for following along. Now go build something with it.