One of my more simplistic projects, Obelisk is a notetaking webapp that has innate smart-searching capabilities. Although it features import/export features to easily move data in between instances, it does not link with an externally hosted database to sync across devices. Everything is in fact stored in local storage, meaning that notes are device-exclusive. However, this does make data fetching and loading far faster than fetching notes from an external database, making it a sleek and efficient electron app instance.

The main motivation for this app was to get some basic web development experience. Obelisk has not only a browser page version, but a web extension and an electron app version, because I wanted to experience all 3 aspects of web development. I also wanted a way to easily search my notes that wasn't just pattern matching as well, so this was something that I found useful for personal use as well.

The main features of obelisk include the smart searching capabilities that matches based on similarity rather than strict pattern matching, as well as the ability to tweak the parameters and weights of this smart search to better define the search itself. Besides this, Obelisk is nothing more than a simple note taking app that can live in your browser as a page or extension, or separately as an electron desktop app.

Really the biggest aspect of development was implementing the smart similarity search. This wasn't some huge efficient super search project, so it's definitely not the most efficient. However, it works pretty well for its intensive purposes!

The way this search works is by tokenizing the search phrase, and then searching every note for those tokens with a few rules:

• A value to a found token in a note is weighed in terms of both frequency, as well as importance. If a value is found multiple times, that weight goes up in a linear fashion. If that weight goes up in importance (i.e. if its in the note title, subject, or content), then that weight goes up in an exponential fashion
• Token sequences are also considered too. The more tokens that are found in successive order that matches with the search phrase, the more the weight increases for that search value in an intensely exponential fashion

Because of this, the search phrase similarity tends to have far stronger weights the more similar the content is to the search phrase in terms of both pattern matching, grammar, and frequency. These weights can be adjusted for personal preference, but this largely follows the trend of the more a note talks about or focuses on the search phrase, the more likely it is that the note is what we are looking for.

Nothing outside of barebones JavaScript, CSS, and HTML was used for this outside of using firebase as a host. It was a pretty simple web project relatively, and was pretty fun to do! I typically hate involving heavy web frameworks such as React and other node packages usually due to how bloated projects can get, so it was really refreshing to work on such a lightweight project.

The biggest takeaway from this mini project for me was the basics of web optimization. To be more specific, using advanced CSS only when applicable. As I wanted to make Obelisk look very sleek and modern, I added a ton of transitions and complex compositor effects to give it a more minimal and modern look. However, this makes it run extremely bad on low-spec devices. Even on phones, the transitions are extremely laggy. In the future, picking and choosing when to use these graphical features is something I need to be much more mindful of. After all, one of the biggest selling points of a webapp is the ability to run it on any device, anywhere.