We require a username when users sign up for my company's product. The username must be unique in our system and cannot be changed after signup. It’s all industry-standard stuff. Unfortunately, another standard part of usernames is how bad our users are at picking them. Many customers have come to us asking if they can change their usernames. The most common screw-up is someone creating a company account and using their name for the username. For example, Ted Smith creates an account for the Acme company with the username tedsmithacme. This becomes a problem when the account is shared, or Ted no longer works for the Acme company.
To reduce the number of bad usernames, I was tasked with using AI to generate suggested usernames for new accounts. I achieved it by taking the user's company name and including it in a prompt to Langchain to generate four username suggestions. Those usernames are then displayed to the customer in the signup form. They are free to select one of the provided names or go ahead with a name they choose. The end goal is fewer regrettable usernames.
Getting Started
In a previous article, I wrote about building a Node.js + express API designed to serve endpoints with AI tools. The username generator was added as another endpoint to the API. This article assumes basic knowledge of Node.js and Express.
Below is the complete code example used. I’ll go over it piece by piece after
import { OpenAI } from "langchain/llms/openai";
import { Router } from "express";
const router = Router();
router.post("/suggest", async (req, res, next) => {
try {
const model = new OpenAI({
temperature: 0,
modelName: "gpt-3.5-turbo",
});
const response = await model.predict(
`Please return 4 usernames for the company: ${req.body.companyName}.
Ensure each username has no spaces, only lowercase characters, and
is a minimum of 6 characters in length.`
);
res.send({ text: response });
} catch (error) {
console.log("Error: ", error);
next(error);
}The first step is creating a model object with the Langchain OpenAI class. I’ve only included the basics for configs. The model variable is a Language Model instance. You can read more about the…