ghostcall opensource work

I have been wanting to contribute to something open source forever but never had the chance to, or in some cases, wasn’t contractually allowed to due to working in healthcare etc. But now that I find myself laid off (AGAIN), I am a free agent and have been looking for where to even begin. I don’t know the etiquette, and most of my really strong interests right now are NOT things I am professionally good at, like working with AI tools and blockchain.

I searched for Slack and Discord communities, joined Discords for any of the tools I tried out recently, since it seemed like I was getting most of my good job leads from the [Gophers Slack][https://gophers.slack.com/join/shared_invite/zt-1vukscera-OjamkAvBRDw~qgPh~q~cxQ#/shared-invite/email] (that’s an invite link if you’re wanting to join, I had to hunt one down). I saw [ghostcall][https://ghostcall.online/] being posted in one of my blockchain Discords and decided to jump on the opportunity since they are having a bit of a “developer challenge” to try to get folks pumped about it and see who should be contributing to help them get their product released in the wild. It’s a p2p network using mobile phones as nodes, pretty novel. The dev challenge is to take their existing video conferencing app (like a p2p Zoom) and add some sort of innovative feature, present your forked code to them and hopefully it’s so good that it makes it into prod! As my first try at this stuff, I figured this would be a good, nothing to lose type of gig since if my code sucks, whatever, they just won’t use it.

Anyway, so I’ve written a feature that is basically a “Sentiment Sidecar” that runs alongside the web app, grabs its audio and video streams and sends them to OpenAI to transcribe to text, then to HuggingFace (not sure if I want to keep this one, probably will try Google for this) for sentiment analysis. It should, in real time, indicate whether it thinks the speaker is joking, being sarcastic, or being neutral. I originally wanted to name this “Autism Sidecar”, because that’s who it is intended for, but maybe Sentiment Sidecar is less OUT THERE, lol. I hadn’t realized there could be pretty bad times for the user if my app gets things wrong, like if it thinks a person is joking when they’re discussing their mother’s death….so I definitely will get my feet wet in the actual fine tuning of models world now.

That code is currently in a private repo as per their request but can share the standalone version I have that should work with any in-browser video app if anybody wants to see it.