Various additions

[micahmo]

This PR makes a few enhancements for the Thunder server.

• Added a workflow to build Docker images. They are hosted right here on GitHub Container Registry. They should be triggered by any push or PR to any branch (so we can easily test things) but the ones based on the main branch will be tagged with latest so that they will automatically work for people pulling with no tag. (https://github.com/thunder-app/thunder_server/pkgs/container/thunder_server)
• Added the ability to customize the Postgres hostname, port, username, password, and database. Added fallbacks in case any env vars are empty.
• Create the Thunder database if it doesn't exist.
  • I'm having a really hard time making this work. The problem comes down to the fact that the Seqeulize connection is instantiated as part of the module import process, so there's no opportunity to do an async task like creating the db. I also couldn't get top-level await statements to work with a ripple effect of upgrading the modules target, etc., etc. Maybe you can help me with this part, or maybe we can just provide instructions for how to create a db.
• Allow the service to start with the APN info, for people only interested in UnifiedPush.
• Ability to generate a test notification (only works with UnifiedPush for now). See #1364.
• Increase fetch timeout as I was having issues talking to ntfy.

I'm going to leave this PR as a draft until I can get the service up and running for myself! EDIT: Actually it looks like drafts aren't configured. 😕

[micahmo]

@hjiangsu Just tagging you on this to let you know I've gotten the server in a good enough place that I think I can daily drive UnifiedPush notifications! We'll see how things go. 😊

[micahmo]

I increased this timeout as it seems like my ntfy server is a little slower than the official one. However, it's concerning to me that a timeout can crash the whole server. Is there any way for us to automatically recover and/or retry? This is where my already non-existent node knowledge falls flat. 😊

[hjiangsu]

Thanks for the additions, especially the docker workflow!

Is there any way for us to automatically recover and/or retry?

There's a few ways to do this! Typically, we can just add a try-catch on the function that caused the error. Then, we can create a queue (if needed) to automatically retry again. Or in our case, we can update the database table record (specifically, the timestamp attribute) so that the notification job catches it again at the next interval.

Could you provide a bit more details on the timeout issue you're encountering? I might be able to find an alternate solution here!

[micahmo]

Typically, we can just add a try-catch on the function that caused the error.

What do you think about putting a try/catch around all of checkNotifications? Since that's sort of the "main thread" of the server, I'd hate for any exception in there to take us down. And yeah, as long as timestamp doesn't get updated until we've done the processing successfully, we should catch it the next time like you said!

Could you provide a bit more details on the timeout issue you're encountering? I might be able to find an alternate solution here!

Sure! I was getting this exception from fetch when pushing to my ntfy server.

TypeError: fetch failed
    at node:internal/deps/undici/undici:12502:13
    at processTicksAndRejections (node:internal/process/task_queues:95:5) {
  [cause]: ConnectTimeoutError: Connect Timeout Error
      at onConnectTimeout (node:internal/deps/undici/undici:6635:28)
      at node:internal/deps/undici/undici:6587:50
      at Immediate._onImmediate (node:internal/deps/undici/undici:6619:13)
      at processImmediate (node:internal/timers:478:21) {
    code: 'UND_ERR_CONNECT_TIMEOUT'
  }
}

[micahmo]

I pushed another change which fixes the sort type when retrieving notifications. Because it wasn't sorted by new and limited to 50, it would miss new notifications and find nothing that was unread (for my "real" account that has >50 replies 😊).

[micahmo]

I pushed another commit which uses the slimmer payload object for reply notifications.