Since my last post I spent a bit of time learning about Large Language Models (LLMs). And while I don’t consider myself to be an expert (possibly because of my own standards), I can say that working with LLMs is very easy.

Back in the days before Open AI released ChatGPT, I remember when natural language to SQL was a difficult sequence to sequence problem. Well, I can say that’s definitely not the case anymore.

In the langchain documents, there is a very clear example of using their library to create a SQL agent.

For those of you unfamiliar with the topic, langchain is a framework for developing LLM applications, and a LLM agent is a specially prompted LLM that takes actions in the environment (e.g. querying a database, searching Google, etc.) to achieve a goal. The ReACT prompt is an example of one these special prompts that turns LLMs into intelligent agents.

Anyways, I took the langchain example and decided to create a SQL agent that answers questions about Pokémon, using the data found in Kaggle. Here’s a link to myGitHub repo with my SQL agent.

As always thanks for reading. In my next post, I may talk about my experiences with the Educative.io platform.