Dlexa | Devpost

Inspiration

Discord calling has been one of our biggest bonding moments as friends. Throughout the years, DIscord has evolved into something more than just a media platform for gamers, whether it be gaming, school, or work meeting, discord has always been a platform for us and many others to use. With that in mind, we decided to take on the challenge of creating a virtual assistant for Discord, Dlexa.

What it does

Dlexa is a virtual assistant triggered through voice activation, allowing the users to use Dlexa in a call as they please. Dlexa grants the user information on top google search results and weather forecasts, as well as, giving users the ability to control music and set reminders through their voice.

How we built it

Dlexa is a Discord bot built using NodeJS and the discord.js library. Assembly AI was used to provide voice control through real-time transcription of conversations. Paired with Discord's API, the ability to transcript a discord call in real-time becomes possible. Many other APIs were used to create these features, such as OpenWeatherMap, SerpAPI (for google searching), google-cloud-APIs(TTS).

The 16-bit dual channel pcm audio stream from discord had to be processed into a mono channel stream for the Assembly AI API.

Both bot.js and vdiscord.js call each other on their own events,

Challenges we ran into

Unfortunately, Discord has recently updated its API structure making it difficult to find ways to extract audio from our voice calls. Although troubling, we eventually solved this problem by using the older version of the API which is still compatible with Discord. Additionally, our project was originally planned to be coded in Python, however, there was no straightforward implementation of audio extraction from voice calls.

Accomplishments that we're proud of

Although Dlexa is far from being a fully finished product, we are amazed at the functionality of the product. Being able to transcript voice calls in real-time and call commands through voice on Discord is something we have never encountered before and we're happy we were able to bring our vision to life. That being said, we believe that the potential for Dlexa is sky-high and we are merely scratching the surface.

What's next for Dlexa

Many features may be added to Dlexa considering the great amount of uses Discord has nowadays. Gaming features such as looking up a user's rank, muting other users/deafening when necessary are just some ideas we have about expanding our project to the gaming side. However, our biggest goal is to develop an algorithm that allows Dlexa to create and organize meeting minutes based on topics and sentiment. We believe this function would open the door to many other important ones in the future as well.