Take-Home Test -- Video API

Description

This video API project uses ffmpeg and ffprobe from the host machine's environment to upload and download video files and metadata from supabase. It also lets the user merge two previously uploaded video files into a single merged file.

Running the project locally

1. Make sure you have ffmpeg and ffprobe installed on your machine, with libx264 for video and libfdk_aac for audio. For example, on MacOS/Linux (using homebrew):

   $ brew tap homebrew-ffmpeg/ffmpeg
   $ brew install homebrew-ffmpeg/ffmpeg/ffmpeg --with-fdk-aac

2. Install node using nvm (or whatever method you prefer). Its always a good idea to click through
   and check scripts for nastiness before piping them directly into bash. A .nvmrc file has been provided in this project to make sure the node version is correct.

   $ curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.39.3/install.sh | bash
   $ cd path/to/project
   $ nvm use

3. Install pnpm and project dependencies

   $ npm install -g pnpm
   $ pnpm install

4. Add a .env file. This file should contain the following properties, used to connect to a properly configured supabase service: SUPABASE_URL and SUPABASE_ANON_KEY.

5. Choose whether you want to run the development or production build version of the project. For Development:

   $ pnpm dev

   For Production Build

   $ pnpm build && pnpm start

Running the project in a container

A Dockerfile has been provided to build the project into a container. Note that we have to build ffmpeg from source because we are using libfdk-aac.

API Docs

Swagger API docs are available at localhost:4000/api-docs (or wherever you map the container url and port to). These let you try out the live api if you want to test it.

Notes about the solution

• Folder based routing:

  Routes are defined by adding folders and files into the router directory. Folders represent path segments, and [bracketed-values] represent dynamic path segments that are injected into the handler as request.params.

  Each route can have middleware provided to it by exporting an array instead of a single Handler function. This is whats happening with multer in the /videos POST endpoint.

• Co-located API docs:

  This project uses swagger-ui and swagger-jsdoc to co-locate the API documentation with the endpoints. Colocation is better for this because it removes a barrier from keeping docs up to date. Nobody likes stale API docs. Swagger was chosen so that we have the tools to live test the API. Live testing using the docs is also a great way to encourage docs to stay up to date.

• Web streams:

  Wherever possible, this project endeavors to use web standards. Fetch was recently added to node.js core, and the version that was added aligns with the whatwg spec. This makes it mildly annoying to interact with the fs, which is written on the older node.js stream standard, but that is the cost of future-proofing.

• Testing:

  Endpoints were all developed using the swagger docs as a TDD mechanism. The swagger docs were written before the handlers and the handlers were built by sending requests from the swagger UI. In the future cypress of playwright should be added to make automated tests.