Every now and then I get a dumb little idea, and too often, I turn those dumb ideas into little web toys. About five years ago, I discovered Markov chains, which in my limited understanding is a deterministic way to guess what would come after some input. A bit like autocomplete for example. If I type, "I like", I'm more likely to type "cats" after that than "yard work". It's fairly complex (see the Wikipedia link above for more details) and perhaps a tiny bit like GenAI. For me, I just think it's neat.

Five years ago I took the excellent titlegen npm package, a list of Cure songs, and built a generator for... well Cure songs: Generating Random Cure Song Titles with Markov Chain. I was thinking about this post and wondered if I could do something with horror movies. Why horror movies? Maybe because it's 200 degrees here and I'm really pining for October and Halloween. Either way, I built it! And here's how I did it.

Getting the Data #

In order for my demo to work, I needed data for the Markov chain. For my data I used the really simple TMDB API. This gives you access to loads of movie and TV data and while I've seen folks use it before, this was the first time I tried it and honestly, I was really impressed with it. To get my data, I hit their discover endpoint with these arguments:

• The horror genre (27), per the docs.
• No video (I believe that means direct-to-video) or adult movies.
• Original language as English

The API can only return 20 results per call but supports paging. I decided to use Cloudflare Workers again because I'm enjoying the platform and knew I could get something up super quick. I want to point out that I'm also using Glitch and Glitch absolutely supports server-side code, I just felt more comfortable doing my server-side code in Cloudflare and my front-end in Glitch.

Ok, so here's the entire worker:

// Horror!
const GENRE = '27';
// cache time of 6 hours
const CACHE = 60 * 60 * 6;

async function getHorrorMovies(key,page=1) {

	let resp = await fetch(`https://api.themoviedb.org/3/discover/movie?include_adult=false&include_video=false&with_genres=${GENRE}&with_original_language=en&page=${page}`, {
		headers: {
			'Authorization':`Bearer ${key}` 
		}
	});
  
  return (await resp.json()).results;
  
}

export default {
	async fetch(request, env, ctx) {

		const APIKEY = env.MOVIEAPI;

		let titles = await env.horrormovies.get('horrormovies');
		if(!titles) {
			let horrorMovies = [];
			const totalPages = 20;
			for(let i=0;i<totalPages;i++) {
			let movies = await getHorrorMovies(APIKEY,i+1);
			horrorMovies = [...horrorMovies, ...movies]
			}
			console.log(`Fetched ${horrorMovies.length} horror movies.`);
			titles = horrorMovies.map(m => m.title);
			await env.horrormovies.put('horrormovies', JSON.stringify(titles), { expirationTtl: CACHE });

		} else titles = JSON.parse(titles);

		return new Response(JSON.stringify(titles), {
			headers: {
				'Access-Control-Allow-Origin': '*',
				'Access-Control-Allow-Methods': 'GET,HEAD,POST,OPTIONS',
				'Content-Type':'application/json;charset=UTF-8'
			}
		});

	},
};

On top, I've got a function that wraps calls to the TMDB API. My worker hits the KV cache (something I'll be blogging about more on Monday, for now, just think of it as a simple key/value caching system) if possible, and if not, grabs 400 movies from the API, filtering out to the just the titles. I cache for 6 hours so the worker can return quicker. Finally, I return the data along with CORS headers so I can use it from another server. You can hit this endpoint here: https://horrormovies.raymondcamden.workers.dev/

Presenting the Data #

For the front end, I used Glitch. I set up my HTML, JavaScript, and CSS there. You can view all the code at the project but I'll focus on the JavaScript.

On page load, I fetch my titles from the Cloudflare Worker, and then use titlegen to initialize the ability to generate titles. As I said, their utility package is super simple. Here's the entire JavaScript file:

let $title, generator, $regenBtn;

document.addEventListener('DOMContentLoaded', async () => {
  
  let titlesReq = await fetch('https://horrormovies.raymondcamden.workers.dev/');
  let titles = await titlesReq.json();

  generator = titlegen.create();
  generator.feed(titles);

  $title = document.querySelector('#title');
  $regenBtn = document.querySelector('#regenBtn');

  doTitle();
  
  $regenBtn.addEventListener('click', doTitle);
});

function doTitle() {
  $title.innerText = generator.next();  
}

As you can see, it's one line to initialize titlegen, one to input the data, and then just running next() to get a new title.

And that's literally it. Play with the full version here:

Obviously, it doesn't always work, but sometimes the "silly" results are funny as hell:

• Final Destination 5: The Addams Family 2
• You Should Have Eyes
• H.P. Lovecraft's Dracula

I hope you've enjoyed this 100% useless bit of code today! Note that the font I used, while excellent, apparently doesn't support numbers.

Photo by Ben Griffiths on Unsplash.