Conversations & Context

A single chat.ask is useful, but the real power is multi-turn conversations. The chat object automatically manages history — every message you send and every response you receive is remembered.

How Context Works

chat = RubyLLM.chat

chat.ask "My name is Alice"
# => "Nice to meet you, Alice!"

chat.ask "What's my name?"
# => "Your name is Alice!"

Each ask sends the entire conversation history to the model. The AI sees every previous exchange naturally.

System Instructions

Set the AI’s behavior with with_instructions:

chat = RubyLLM.chat
chat.with_instructions "You are a pirate. Always respond in pirate speak."
chat.ask "What is Ruby?"
# => "Arrr! Ruby be a fine programming language, matey! ..."

You can also append to existing instructions:

chat.with_instructions "Keep responses under 2 sentences.", append: true

Your Task

Open conversations.rb and build a multi-turn conversation:

1. Create a chat and set a system instruction with with_instructions
2. Ask multiple follow-up questions — the AI should reference previous answers
3. Print the conversation history with chat.messages

chat = RubyLLM.chat
chat.with_instructions("You are a friendly Ruby expert. Keep answers concise — under 3 sentences.")
chat.with_temperature(0.7)

response = chat.ask("What are blocks in Ruby?")
puts "A: #{response.content}"

response = chat.ask("Can you show me a simple example?")
puts "A: #{response.content}"

response = chat.ask("How is that different from a Proc?")
puts "A: #{response.content}"

# Print full history
chat.messages.each do |msg|
  puts "[#{msg.role}] #{msg.content.to_s.lines.first&.strip}"
end

Run it:

$ ruby conversations.rb

Accessing History

chat.messages returns every message in the conversation:

chat.messages.each do |msg|
  puts "[#{msg.role}] #{msg.content}"
end
# [system] You are a friendly Ruby expert...
# [user] What are blocks?
# [assistant] Blocks are anonymous functions...

Each message has a role (:system, :user, or :assistant) and content.

Temperature Control

# Low — factual, consistent answers
chat = RubyLLM.chat.with_temperature(0.2)

# High — creative, varied responses
chat = RubyLLM.chat.with_temperature(0.9)

Temperature ranges from 0 (deterministic) to 1+ (creative). Default is usually around 0.7.