Shunsuke Mori (kokuyou)
Shunsuke Mori (kokuyou)
  • @kokuyouwind

Ruby on Rails engineer. He works at Leaner Technologies inc. as a member of the technical infrastructure team and surveys LLMs as a part of his technical research. He is also interested in type systems and is developing RBS Goose for RBS type inference.

Let's use LLMs from Ruby 〜 Refine RBS type definitions using

Large Language Models (LLMs) have evolved rapidly over the past few years and there have already been many practical applications in a variety of fields. Though Python is recognized as a de facto standard language to use LLMs, we are Rubyists so we would like to explore how we use LLMs in the Ruby language.

In this session, we'll cover Langchain.rb as a way to use LLMs in the Ruby language and share examples of managing prompts and using various LLMS. The development of RBS Goose, which is a tool to refine RBS type definitions, is a good practical example. We’ll share how it works including prompt construction, as well as the tips for LLMs development and our evaluation of its performance, including accuracy and speed, at the combination of various prompts, Few-Shot examples, and LLMs.