Raw Words

Daily broadsheet generator

8 min read

3/30/26

Do I want to use React and a SPA for this? It probably makes sense for me to just do that since it is what I know. I am increasingly feeling like that approach is just kind of silly. I think the main broadsheet for this project could make sense as a React application. The entire site though?

On the other hand, maybe I should just use a React SPA as practice for work anyway. Despite kinda thinking it feels like overkill.

3/26/26

I got part way through setting up auth users with better-auth. I need to finish setting that up.

3/25/26

Open links:

3/16/25

Next stuff?

  • I think I should make some endpoints…
  • But before I get to far, I think I should consider trpc.

Okay, querying data and dealing with form validation nicely is a PITA in React. It’s overly verbose for the things I want to write.

  • I think this is why people like react-query, we use this at work for data fetching
  • react-hook-form I used to like, but its pretty unfun to use at work… i think maybe the library is old and brittle now? i think we’re also incredibly misusing the library too. workflows just should not be a form…

3/15/26

  • I set up drizzle-kit. It just needed a simple config file and a connection to the DB.
  • I’m using push w/ drizzle-kit right now. It seems like that should be fine for dev cases.
  • I might want to switch to a generate and migrate flow once the application is actually running and I want to be sure of how my data works.
  • I had to update my DB indices to reference the tables they’re referencing. This seems to be convention. I think index names need to be unique.
  • I had to setup dotenv-expand so that I could use the db user and pw template vars in my db url variable.

3/13/26

Took some notes on Docker relevant to this project.

3/8/26

Trying to get the env up running

  • I think I have a docker compose and dockerfile created as needed to run on my NAS
  • The docker compose file does not run on my local computer, the reference to a volume that exists on my NAS does not exist on my local machine
    • You can run docker compose up and docker compose down to run everything locally when it is working

I’m also just using ts-node right now to skip dealing with compilation. I understand this is slow and I should add a build step.

  • I could consider trying out deno or bun or something like that, but I’m not really interested in bleeding edge stuff…
  • I do really like bun when it works, but then sometimes unexpectedly there are places without parity. I think this would be fine for a power user, but that aint me dog.

What’s the deal with NODE_ENVs? I need to manage my .env files in local dev as well.

To run:

  • I need to ssh into NAS, git checkout the repo on the NAS, pull it down
  • Copy paste docker compose contents into Container Manager on NAS
  • Set up ENV VARS as expected in the NAS
  • Start the project on NAS

Then, I have to actually build the rest of the application.

  • I have a rough table schema mocked up with drizzle.
  • I want to figure that out and how to use drizzle-kit

Then, I need to actually start making a UI

NOTE: I do wonder if there’s a way to pull in my miniflux feeds, that is also running on my NAS, and then use a real feed aggregator instead of making my own version of that too for this project. I think I should create a simple “ENTER URL” form, and then work on “integrations” for other feed providers, like miniflux.

End of the day check in. I have learned a lot about a lot of things.

  • Dockerfile to build my express app
  • Docker compose starts express and db
    • I had to add ports to both of my services, I’m not sure if I need/want to do this in real life… I think this is exposing my app and db to everyone who is on my network?
    • That doesn’t seem ideal, I do want to expose the web server on a port, I suppose, but it seems strange to let anyone but my web app look at my DB? Maybe that’s just how things work though…
  • I can run just one of my services, the db with docker compuse up <service>
  • I’ve created a few package commands
    • start - this is what the docker file runs to start the app
    • dev - this is what I use to work on the app
      • I tried to just use docker compose up for this, but it didn’t really work how I wanted…
        • It seems like there’s probably some way to get this going, but I gave up
        • You have to keep rebuilding the app image to get new files after you build them, that’s a PITA, but also it is slow and rebuilding a bunch of node things for no good reason, when only a couple of my files changed
      • Instead, I’ve created dev commands for fe, app, and db
        • fe is using vite as a dev server
        • app is using tsx watch
        • db is just using docker compose up db since we don’t actually need to change anything
    • build - this is a step that we do inside the app’s Dockerfile to pre-process anything we have to do before starting
      • this is just building the front end files with vite right now
  • I switched from ts-node to tsx, it seems like the latter is more modern. It has watch built in. And importantly I hit some sort of common.js or ESM related error message, and then just switched to tsx and it went away. I think it might have been some sort of import.meta.filename thing tsx does for you to be nice.
  • I am able to connect to my local DB using SQL Tools extension in VS Code, it seems nice thus far and I think is free/OSS?
  • I installed postegres@18 via brew to my machine to test connecting to the DB to verify it was working. I had to add psql to my PATH to get it to work after installation.

Next time:

  • I need to get set up with drizzle kit, I think. My database is totally empty, as expected.
  • I have a users endpoint that gets no data because nothing is there to get!
  • I need to do my initial migration.
  • It seems like there are two strategies people use here that I’m interested in:
    • https://orm.drizzle.team/docs/migrations
    • #2 on that list is drizzle-kit push
    • #3 is drizzle-kit generate and drizzle-kit migrate - i’m more familiar with this approach and the other one seems a bit suspicious? Too good to be true? I dunno
  • When I ctrl+c my npm run dev command everything doesn’t clean up how I want it to
    • I need to run docker compose down to actually stop all the containers

3/7/26

I had a conversation today about how a friend wants a service/software that generates a daily digest of their RSS feed instead of an actual feed.

This got me thinking about my Creating my own social media or information or RSS feeds thoughts.

I think Feeds are addictive. Maybe not necessarily, but I am constantly refreshing them hoping for scraps of entertainment through out my day. Even feeds where nothing interesting is likely to show up.

I like the idea of instead of trying to make a curated RSS feed, instead trying to create a finite digest. Once you’re done, you’re done. I’m not sure if I want it daily or weekly or what. But I think it would be cool if it was finite.

This more mirrors my relationship with NYT games or other dailies. Those can be a bit pernicious, especially if they track streaks. But, generally, I feel like these things are a good use of my time.

I feel fine when I miss content in daily puzzles. If I’m half finished and the day flips over, whatever. It’s a little annoying, but I don’t mind.

I’m not sure how I feel about missing some content. There are some creators who I do want to see all their publications.

I like the idea of having broader controls over the broadsheet I generate. Just keep dumping cool sites in and see what happens. The generator will keep it all contained to a manageable scope.

What do you need to build this?

  • A site that serves the broadsheet for mobile and desktop. I think I’ll care most about tuning this experience.
  • Backend
    • Cron job to generate the daily broadsheet
      • This needs to read an RSS feed
    • Database to store the information we need to construct the broadsheet
    • Database needs to collect RSS feeds
  • I need an input field where I can submit an RSS feed… It would be cool if I could use a different RSS feed manager and just send its output over to the generator… I’m not sure if you can “collate” feeds like this.
    • I imagine you could do so
    • I guess the real question is, do normal feed accumulation apps actually expose a feed of what you’ve fed into them?