But I can I can get into this or if you want me to go? It's only gonna take maybe 3 minutes. Yeah. You can get into it. Okay. Yeah. I love this little light. House. Yeah. I can see. I'm going to get up. How much we gotta do for the house because I basically can't see this. Oh, okay. I don't know if somebody Oh, yeah. You got rescheduled to Saturday. Oh, really? 5 PM. So, yeah, that's basically super high level design, but Oh, so beautiful. Harder to based on it harder to put me, but, you have your demand of update, upload, delete, and this is basically this chunk of the upload, edit, and delete side of things. You'll have your application server here, like some easy 2 or wall of them, on auto scaling. Video MP 4 file goes into a, optic, store that's contains the raw, video format data, and it gets added to a messaging queue. That messaging queue then basically pulls all the video uploading processes together, help a ton in terms of how much compressors you need, encoding process. That's gonna help a ton in terms of how much compressors you need, encoding processes. Because that's you know, you can only encode 1 literally 1 compressor. It only takes 1 video at a time. The vid the video appears pre encoded from the device. So when at all possible, it would be good to keep the encoded block of video. I guess you need to decode it for analysis at least once. I'd imagine I'd I'd imagine y'all would have more, clarity on that Yeah. Part here because this was supposed to be Okay. I I I can I can fill stuff in? Yeah. But in general, the process yeah. I mean, it's been a much if you let them this is basically all, you know, auto scaling up and down. And then here, this is, like, man managing capacity, then you store, you know, the encoded object into the objects over here. And both MongoDB, being actually the best, nearly the best database option for storing all this stuff. It's extremely read heavy. Like, it's just constantly reading everything, everything, everything. That's where the watching is served, Longitude and Redis, will come in. So anytime you're watching a video and stuff like that, you'll be storing well, if you guys wanna go even more luxury, which typically you should be doing is, having a CDN Uh-huh. A network to deploy all this, audio, video, metadata, crap like that. You can get into that later, but, you know, watching a survey, you're definitely gonna have a cache to make sure a lot of this the popular videos to people having on their on their profiles. They get loaded extremely fast. And, of course, you know, the encoded videos, they go here, and Mongo just points exactly to which one is needed. We send those chunks, and these basically are at a, they're not a it's not gonna be using, like, JRPC, like, with with basically OpenAI's real time chat uses. It's really just standard API call chunks, and we just send chunks directly where the pointer is. And that's it. We just send those chunks. We don't do anything else. We're not actually streaming anything. So those chunks are, like, maybe, let's say, 10 second chunks. We send them 5, 10 second chunks over an API call. And the front end guy will basically say, hey. Start here and then start sending me chunks to get to this buffer. And then as the vehicle goes on, the front end gives us more chunks, chunk, chunks, chunks. That's exactly how actually YouTube works. They do their they do their their stuff over QUIC, HTTP 3 quick protocol, but, HTTP 2 is still just fine. And we're not I mean, you could do this other ways where you have a constant WebSocket stream and management. That's going to be a lot more stress on, other designs that I've seen. So generally speaking, this is kinda a clone of what YouTube is. If it works for YouTube, it works for us. Very simple. I'd imagine, very simple understanding of serving and then upload, edit, and delete. Like, I just went in and storing it. It's a super fast I can code this up in in not too much time. Mhmm. And then, generally, you know, I can we can fill in what y'all have for your specific devices and all the weird nuances that come around it because this is still just extremely high level. Yeah. Yeah. Yeah. One one one good thing to have would be the ability to watch a streaming preview of the feed from the device from the back end just to, like, be able to, like, set the device in the right spot and position it. Because I know you're gonna have, what I don't know how the encoding sorry. The metadata behind the video. I mean, that's all stored here. All metadata you can ever imagine gets stored in the document DB. But, like, the geolocation, perhaps I know you guys have bad stuff. I am unsure whether or not geolocation data gets stored in the video into here or or that gets geolocations get stored somewhere else. It's not a part of the video processing, you know, back end. Because I kinda wanna have our separation concerns of every single, you know, main item, like like audio, video, image. And do you