This project is a collection of conversation samples between professional health coaches and their clients. This data can be used with Large Language Models (LLMs) as one- or many-shot examples, or as data for fine-tuning, to replicate the behavior and performance of a qualified health coach.
All conversations were synthetically generated by one or more LLMs, and passed through multiple automated reviews to ensure that conversations adhere to coaching best practices and did not contain irrelevant or inappropriate content.
100 conversations as arrays of messages in JSON format, stored in the /conversations folder. The topic-index.md file groups the JSON files by category and provides a brief description of each conversation for easy reference.
- The client is labeled as
user. - The coach is labeled as
assistant. - Conversations begin with a request by the
userand end with a response from theassistant.
The data provided in this project has been reviewed, but may still contain statements that some may find offensive. This project is free to use (see LICENSE), but should be used responsibly and for research and testing purposes only.