GPT Todo List

Take a ToDoList class and chain it to GPT-3.5 by passing the methods for interacting with the list as LangChain tools or OpenAI function calls.

Implementing a to-do list is an important rite of passage when learning a new application development framework, so it woud be fitting to make one with GPT and Langchain.

In this demo, we take a ToDoList class we’ve prepared in advanced (well … we had ChatGPT prepare it for us, of course), and we chain it to GPT-3.5 by passing the methods for interacting with the list as Langchain tools.

When interacting with the chat, it figures how to use the tools to read and modify the to-do list, thus becoming our UI.

This demo is a bit limited - as you’ll see, the instructions we give the chat are quite specific and there is not a lot of planning done by the chatbot, beyond a couple of steps. Still it’s pretty cool to be able to chat to a to-do list and have an AI agent modify it on our behalf. And there’s a challenge for future iterations: to make the bot acceppt more absrtact requests and do more of the planning itself.

Another implementation relies on the native Open AI function calling API, without relying on LangChain for orchestration. It works well for simple requests, but fails in complex interactions that require chaining multiple function calls.