From string to sing: How ChatGPT can write your rails' Active Record code

Let's face it, fellow Rails devs: a lot of what we do is just basic CRUD. We spend most of our days translating user requests into queries that hit our database. And let's be honest, we're not animals - we're not writing raw SQL. No, we dress that baby up with shiny Active Record calls to make it look pretty. But what if I told you there's an even better way to do this? Yes, my friends, we're talking about using the power of LLM (that's large language models, for those not in the know) to easily take our user queries and turn them into Active Record calls. And today, we're going to look at three different methods for doing just that.

The setup

To demonstrate how to use the power of the LLM to enhance your app, we will be working with a simple Rails app called the rails-railx-tailwind starter kit. This starter kit consists of User and Article models, with a one-to-many association between them. The source code for this app is available on GitHub. Let's dive in!

Here's the code for the User and Article models:

class User < ApplicationRecord
  has_many :articles, dependent: :destroy
end

class Article < ApplicationRecord
  belongs_to :user
end

Method 1: The way of the gem

It's time to dive into the first method of bringing some AI goodness into our Rails apps! And we're not talking about sprinkling some machine learning fairy dust on top - we're talking real, honest-to-goodness artificial intelligence powered by ChatGPT. And the best part? We don't have to do all the heavy lifting ourselves, because an open-source gem called Boxcars makes it a snap. Just add the gem, make a quick call, and watch as Boxcars takes care of all the parsing and handling the AI magic for you. It's like having your own personal AI assistant without having to pay them a salary. Check it out:

require 'boxcars'

ar_boxcar = Boxcars::ActiveRecord.new(code_only: true)

puts ar_boxcar.conduct("How many articles has John written?").to_h[:code]

3 simple lines of code, and our work is done.

We'll get back something like:

User.where(name: "John").joins(:articles).count

Method 2: DIY with a little elbow grease.

All you DIY enthusiasts out there, get ready to flex those coding muscles! Our next method for using ChatGPT to write ActiveRecord queries in Rails requires a little elbow grease, but don't worry - it's nothing you can't handle. We're going to roll up our sleeves and make the API call to OpenAI ourselves, manage the prompt, and parse the output. No Boxcars to hold our hands this time, just pure unadulterated AI power. Let's take a look at the code:

require 'openai'

client = OpenAI::Client.new(access_token: "your-token")

PROMPT = <<~IPT

Given the following Active Record models:

class User < ApplicationRecord

  has_many :articles, dependent: :destroy

end

class Article < ApplicationRecord

  belongs_to :user

end

Please write me an Active Record query that returns the answer to the following question.
Only reply with the Active Record query

Question: How many articles has John written?

IPT

response = client.chat(
    parameters: {
        model: "gpt-3.5-turbo", # Required.
        messages: [{ role: "user", content: PROMPT}], # Required.
        temperature: 0.7,
    })
puts response.dig("choices", 0, "message", "content")

We'll get back something like: User.where(name: "John").joins(:articles).count

So here's the deal: we're using the openai Ruby gem to make the API call to ChatGPT. First, we set our OpenAI access token- don't forget to replace 'your-token' with your actual key, or else things are going to break. Next, we provide a prompt for ChatGPT and set a temperature of 0.7

Once we make the API call, we'll get a response that contains the answer. We'll have our shiny new ActiveRecord query ready to use. Go ahead, give it a spin!

Method 3: Smooth Talkers

Are you a smooth talker? Then this method is for you. Forget about APIs and coding; we're going to have a one-on-one conversation with ChatGPT. It's like talking to a friend; only this friend is an LLM with impressive programming skills.

Here's how it works: head over to https://chat.openai.com/chat, type in the following prompt, and let your conversational skills do the rest:

Given the following Active Record models:

class User < ApplicationRecord

  has_many :articles, dependent: :destroy

end

class Article < ApplicationRecord

  belongs_to :user

end

Please write me an Active Record query that returns the answer to the following question.

Question: How many articles has John written?

After a bit of chitchat, ChatGPT will give you the goods:

So put on your smooth-talking hat and get ready to charm ChatGPT into giving you the Active Record query you need.

In conclusion, adding AI features to Rails apps is becoming more accessible with tools like Boxcars and openai gem. We also explored a manual approach to interacting with the LLM by chatting directly with ChatGPT on the OpenAI website. With these three methods at our disposal, we can leverage the power of AI to help us write Active Record queries more efficiently and effectively. By combining the domain-specific knowledge of our apps with the natural language processing capabilities of the LLM, we can unlock new possibilities in Rails development. So go ahead, try it, and see what kind of Active Record queries you can come up with!