Hey, Selfbook devs! This is µv, a barebones movie recommendation engine based on the prompt you guys gave me.
This repo uses conventional commits with bitbucket/github emojis. You'll find most of the relevant code under packages including a few tests.
So there was a little ambiguity on how the prompt defines "application". It could be interpreted as basic CLI with a file input or as a standalone web application. As such I'm going to attempt to satisfy both in the form of this monorepo (UPDATE: Just the CLI).
Update: After getting the repo ready I slept on it and worked through how I'd like to approach the computational end of the app. I have a little working knowledge of graph data structures and think a weighted + undirected graph fits the requirements pretty well. Assuming I have the time for it today Given the opportunity to expand the scope, I'd implement it like so:
- Create a "random" suggestion algorithm
- Compute raw popularity stats for movie data
- Attach to CLI
- Create a minimal
WeightedGraphclass and implement it as an adjacency list (optimized for looser relationships like this) - Use that graph class for:
-
UserGraphfor computing user overlap -
MovieGraphfor computing movie similarity/correlation
-
- Precompute these values
- Attach to CLI
- Crate basic web app implementation
- Use "random" algorithm as a fallback?
- Polish web app UI
NOTE: Might require Node 18.x!
# This project uses pnpm's workspace feature, it's very lean!
pnpm i
pnpm build
# Use the cli
npm link
mu-v --help
mu-v --file ./data/movies.json --user 27 --count 6
# Also works with shorthand arguments
mu-v -f ./data/movies.json -u 17
# this might work too?
# npx run mu-v 5 data/movies.json
# Also:
# pnpm coverage
# pnpm testFolder structure looks as follows:
app- A web app, framework yet to be chosendata- Static JSON and/or mocked database APIpackagescli- terminal clishared- functions and types shared between both builds