Stack overview

Now that I have been working on the project for several weeks and the technologies being used are starting to settle down, I thought I would share some insights on what I am using — and perhaps also the reasoning behind these choices.

Authentication :: google

Well, the decision here was kind of easy. I wanted to go with some general OAuth flow and not implement anything on my own. As I am on Google (like many others), I decided to start with it. The way I use it will also enable me to add several other providers later on if they are required.

Database :: AWS DynamoDB

The project does not need a lot of database technology. At least as of now. Some storage for user data and the bots. As I wanted to be dynamic on the schema side, I decided to go for a NoSQL solution. AWS DynamoDB is a very nice offering for a quick start.

Caching :: Redis + AWS s3

Reposting requires some fast caching to ensure no double-postings happen. For this, I use a local Redis host. To keep the Redis instance simple, I additionally put the keys into an S3 bucket on AWS as a persistent second layer. This way, I can restart the whole application without losing the cache.

Backend :: fastapi

As I have been working with FastAPI for the last 3 years, this was a natural choice for a simple yet robust solution for all the APIs.

To schedule the required jobs, I use apscheduler. As with FastAPI, I just love it for its simplicity.

One thing I added only very recently is fastapi-events. It provides means to dispatch events while handling requests, which will then be handled after returning the responses. A similar idea to FastAPI’s native Background Tasks, but way more flexible.

Frontend (app) :: Vue + Vuestic

Initially, I wanted to do the frontend in React… But while hiking, @mademyday convinced me - simply by being very opinionated about the topic ^😉 - to give Vue a try. Well, and I did.

And, as the section headline might have already given away, I kept going with it. It just feels way more natural to create frontends with it. Clear and simple.

As I did not want to create all the frontend components of the app on my own, I also checked for a frontend component library. After reddit-ing for a few hours and some simple tests, I settled for Vuestic UI and did not regret it so far.

Frontend (charts) :: chart.js

Although I'm usually not much of a frontend developer, I put more passion than expected into the first version of the metric charts.

The main library I use is chart.js - or, to be more precise, vue-chartjs. On top of it, I use the following extensions/plugins:

• chartjs-plugin-gradient for nice gradients that most of us may know from Grafana. I just had to have some gradients in my area charts 😉

• chartjs-plugin-annotation - the name gives it away, a very flexible tool for chart annotations. I use it for the big overlay numbers that show the “totals” on top of my charts.

• chartjs-scale-timestack for the time axis. It does a fantastic job of rendering nice values automatically on different time-scales.

Frontend (web) :: Jinja2

As of now, I have decided to do all “landing pages” with basic HTML. This means I will use FastAPI + Jinja2 templating here. Again, mainly because I am used to it and not much frontend-fanciness (other than styling) is required. If this changes at some point, I will switch over to some Vue-based solution…

One thing I want to mention, just because it made my life much easier: I included arel, which provides a nice page-autoreload-solution. The only downside is that I could not manage to make it detect changes on my template files, which are mounted into a Docker container. This is why I use uvicorn’s --reload functionality for now. But at least I do not have to reload the page every time I change the code.

CSS :: tailwind

This was also an easy decision. Firstly, it is easy to use, and I simply love the way one can create new classes using the @apply tool. Secondly, it is also easy to integrate it into Vuestic, so, in the end, I might even be able to create a central theme for both implementations.

Infrastructure :: k8s, argocd, prometheus, grafana

All backend code is run on a Kubernetes cluster, which is provided, maintained, and hosted by the one and only DrPsychick.

Full disclosure: I just love Grafana dashboards. I use them as my daily dose of motivation. This way, I can always see that something is happening on my projects.

Mails :: MailJet

Not much to say here, I think. I send all transactional emails via MailJet’s APIs. A simple web frontend to create the emails. And an even easier API to actually send them. No pain in trying to resend etc. MailJet got me covered (for now).

Any Opinions?

As you may have realized, I tend to try things that are recommended by highly opinionated people. So, send me your opinions and suggestions. What am I doing completely wrong in your opinion? Which is the one API/library I missed?

Anything else you want to know/share/contribute? Just ping me on Bluesky, Reddit, or GitHub. Looking forward to meet you there.