ChatGPT Proves How Good My Music Taste Is

Like you, I’ve also been anxiously awaiting the Sade x Meshuggah tour. But I just don’t think it’s gonna happen.

What you see above is the result of entirely too many hours interacting with ChatGPT to display my nostalgia-based vinyl collection.

You might (not) be asking yourself, how’d you accomplish such a feat? Simple, really. I have entirely too much free time. But it does involve an interesting tech stack, so if you’re into that kind of thing, read on!

Enter Discogs

Discogs is basically the Untappd of the vinyl world. Meaning, you can add a record to your “collection” and share a link with your friends to compare records, trade records, etc.

Long story short, the data just kind of exists. I thought to myself, wouldn’t it be cool if I could query that data somehow, and maybe display it in an interesting way?

Enter ChatGPT

Full disclosure, I am pretty bad at reading API documentation most of the time. ChatGPT basically lets me skip that hastle. So the first thing I looked up was:

import requests
import json

# Your Discogs username and API key
username = 'your_username'
api_key = 'your_api_key'

# The ID of the folder containing your collection
folder_id = 0

# Base URL for Discogs API
base_url = 'https://api.discogs.com'

# Endpoint for retrieving collection releases
endpoint = f'/users/{username}/collection/folders/{folder_id}/releases'

# Set up authentication headers
headers = {'User-Agent': 'MyDiscogsClient/1.0', 'Authorization': f'Discogs token={api_key}'}

# Make request to API
response = requests.get(base_url + endpoint, headers=headers)

# Convert response to JSON and output to file
with open('my_collection.json', 'w') as f:
    json.dump(response.json(), f)

I simply saved that locally as get_collection.py, created an API key in my Discogs account, and hardcoded my credentials without a care in the world. I then executed the script and verified the output to the newly created my_collection.json file. So far, so good!

I needed to parse out the actual URLs of the albums I own so I could then save them and arrange them in a nice grid for the world to see. That said, ChatGPT is rather verbose in its output so I will just be pasting in the relevant code snippets from here on out.

# Extract album thumbnails from each release
thumbnails = []
for release in response.json()['releases']:
    # Check if release has an image and add it to the list of thumbnails
    if release['basic_information']['thumb']:
        thumbnails.append(release['basic_information']['thumb'])

After a bit of jq inspection of my_collection.json I found that the preferred thumbnail was actually release['basic_information']['cover_image'] which was a simple enough edit. If you read on you will notice that not all of ChatGPT’s results are direct answers, but they get you close enough. Which brings me to an issue with the above code; pagination (typo to follow).

# Make request to API and retrieve all pages of results
page = 1
thumbnails = []
while True:
    response = requests.get(base_url + endpoint, headers=headers, params={'page': page})
    response_data = response.json()
    # Extract album thumbnails from each release
    for release in response_data['releases']:
        # Check if release has an image and add it to the list of thumbnails
        if release['basic_information']['cover_image']:
            thumbnails.append(release['basic_information']['cover_image'])
    # Check if there are more pages of results
    if 'next' in response.links:
        page += 1
    else:
        break

The above code resolved the pagination issue, and now I could retrieve all of my collection (79 records at the time of writing). Now that I had a list of thumbnails called thumbnails it was trivial to iterate through them and perform a wget operation on each.

import os
import subprocess

# Set the URL and file name for the file you want to download
url = 'https://example.com/myfile.txt'
filename = 'myfile.txt'

# Set the directory where you want to check for the file
directory = '/path/to/directory'

# Check if the file exists in the directory
file_path = os.path.join(directory, filename)
if os.path.isfile(file_path):
    print(f"{filename} already exists in {directory}")
else:
    # Download the file using wget
    print(f"Downloading {filename} from {url} to {directory}")
    subprocess.run(['wget', '-P', directory, url])

With a bit of Frankenstein code with the above snippet and what we saw previously, I was able to come up with this:

for thumbnail in thumbnails:
    filename = thumbnail.split("/")[-1]
    file_path = f"{directory}{filename}"
    if os.path.isfile(file_path):
        print(f"{filename} already exists in {directory}")
    else:
        # Download the file using wget
        print(f"Downloading {filename} from {thumbnail} to {directory}")
        subprocess.run(['wget', '-P', directory, thumbnail])

This resulted in a collection of 78 photos in my thumbs/ directory. Interestingly, this was one less than my total collection… I believe a single record was missing the required release['basic_information']['cover_image'] and thus no thumbnail was downloaded.

Now was the time to arrange everything in a nice grid using some semblance of html/javascript/css. My skills in front end development are nonexistent, so…

Enter More ChatGPT

<!DOCTYPE html>
<html>
<head>
  <title>Image Grid Example</title>
  <style>
    .grid-container {
      display: grid;
      grid-template-columns: repeat(auto-fit, minmax(200px, 1fr));
      grid-gap: 10px;
    }
  </style>
</head>
<body>
  <div class="grid-container">
  <?php
  $dir = 'path/to/images/';
  $files = scandir($dir);
  foreach ($files as $file) {
    if (is_file($dir . $file)) {
      echo '<img src="' . $dir . $file . '">';
    }
  }
  ?>
  </div>
</body>
</html>

Oh no, PHP. If there’s anything I like less than JavaScript, it’s PHP.

• code response

In order to prevent this post from getting too long, I had quite a bit of back and forth with ChatGPT from here on out in terms of the front end design. Here are a few of my questions before we cut to the final layout:

• code response

• code response

• code response

• code response

After modifying my index.html file I placed it in my project directory and launched a simple python web server: python3 -m http.server

I could then load 127.0.0.1:8000/index.html in my browser, greeted with my incredible record collection in its full glory:

Now came the part in my project to share the above with the world. Thankfully I had a barren wasteland of an cloud account at the ready…

Enter AWS

Without going into insane detail on the infrastructure, I knew that this could be something like:

1. A Lambda function that would:
   • Grab my collection data using the Discogs API
   • Read an S3 bucket subdirectory called thumbs/ and check for missing thumbnails
   • Download missing thumbnails to /tmp/ and then upload them to the S3 bucket thumbs/
2. The S3 bucket mentioned above would host index.html
3. The trigger for the lambda function would be a CloudWatch Event running on a cron schedule
4. All of the above but in terraform because why not

That said, I used ChatGPT for all of it…and it works!

Live demo here, for now

Note: I realized mouseover/hover is kind of weird no mobile so I just modified the above to pick a random album and pulse, repeating infinitely.

In Conclusion

If you’ve used ChatGPT for any code-related endeavors you most likely arrived at a similar conclusion: ChatGPT is an imperfect tool that makes finding the answer a much easier task. Just a few months ago I’d be searching similar terms using Google and perusing the first 5-10 results; now I just type the question into a prompt and do a bit of followup when required.

Thank you for reading 😎

Check out the code on GitHub