Turn | HyperKit.ai

Definition

A turn is a single exchange in a multi-turn conversation between a user and an LLM. Each turn typically consists of a user message followed by an assistant response, forming the basic unit of conversational interaction.

Multi-turn conversations allow for iterative refinement, follow-up questions, and complex task completion that would be difficult in a single exchange.

Key Concepts

Turn pair: A user message and the corresponding assistant response
Multi-turn conversation: A series of turns building on each other
Turn-taking: The alternating pattern of user and assistant messages
Turn count: Often used to measure conversation length and complexity

Examples

Single Turn

One-Shot Interaction

Turn 1:
  User: "What is 2 + 2?"
  Assistant: "2 + 2 equals 4."

(Complete interaction in one turn)

A simple question-answer exchange that completes in a single turn.

Multi-Turn

Iterative Refinement

Turn 1:
  User: "Write a function to sort a list."
  Assistant: "def sort_list(lst): return sorted(lst)"

Turn 2:
  User: "Make it sort in descending order."
  Assistant: "def sort_list(lst): return sorted(lst, reverse=True)"

Turn 3:
  User: "Add type hints."
  Assistant: "def sort_list(lst: list) -> list: ..."

Each turn builds on the previous, iteratively refining the solution.

Interactive Exercise

✎

Count the Turns

How many turns are in this conversation?

User: "Hi!"
Assistant: "Hello! How can I help?"
User: "What's the weather?"
Assistant: "I don't have weather data."
User: "Thanks anyway."
Assistant: "You're welcome!"

Pro Tips

Keep turn count in mind when managing context window limits
Each turn adds to token usage—summarize earlier turns for long conversations
Design prompts to minimize turns needed for task completion
Use multi-turn for complex tasks requiring iteration or clarification

Definition

Key Concepts

Examples

Interactive Exercise

Related Terms