 Okay. Welcome to Minecraft DevSync, October 14th, 2020. We're coming up on the end of our sprint here and just doing a check-in to make sure we're on track. We'll find out if there's anything we can do to get us back on track. We'll just start with who gets early in the morning for you. Yes, perfect. I posted the precise model that Ken made out to the community yesterday. I think it might take a couple of days to get many responses back because it requires a little bit of, you know, editing your configuration and things to test it out. I was testing a bit more myself and realized that it works really well for me and Beck. It works really well on the Mark 1 and doesn't work at all on the Mark 2, which I think kind of highlights the, well, adds weight to the possibility that we provide different models for different hardware, you know, different audio profiles that are coming through. Anyway, I presume it's because it's just been trained on samples that we collected through the Mark 1s, which have completely different audio to a nice clean feed from a microwave. A few fixes for skills that have been clogging up the, well, just a few fixes for skills and then running into some issues with the old integration test runner that I'm working through some of its intent related and some of its, I haven't figured out yet. And then I've got some benchmarking stuff from Aditya as well on Mark 2 images, including the, what is now known as, my brain's escaping me. It's exciting in the morning. It was formerly known as Minecraft OS anyway, but is now is now known as something else. Yeah, I'm doing some write up in Confluence on different options there and putting all the things in there so everyone can have a look. Anyway, that's probably enough for me now. Okay, thanks. Derek. Alright. Yeah, mostly I'm working on a rev 2 of the SJ 240 fdm design. Kind of impromptu meeting with Kevin because he was having some issues that he wanted to talk about with putting his together and I'd already kind of discovered several issues myself. Mostly around the, well, one was the holes for the grill I think we talked about last time, and then some of the mounting holes and how it was oriented for printing and such. So I've been working on an update for that. The other thing we talked a little bit about was, I had a company reach out to us. They would use before for, well, we had to actually use it, but we'd quoted through them before to do kind of low quantity 3d printing and they just have a new program to transition to tooling. And so they reached back out to us. So I talked to Kevin a little bit about to a little bit about that but also about this other company that we looked at macro fad has a box build as something that they can do now so we're asking to throw together. We hadn't actually quoted with them, I think only mainly because of the lead times it was kind of on our short list. So I want to get a call over them maybe next week and talk to them about that because they can apparently do the PCB assembly and also do final product assembly they don't do the plastics we got to ship plastics that final product assembly and testing. So, that's really interesting. That's been been basically what I'm up to some hoping so Kevin's like okay well I'm gonna wait till you get these updates done before I continue on my prototyping his printing. So I'm hoping to have that done tomorrow or Friday. Let's see. I think everything else is I've got the camera stuffs in in the sprint but I haven't actually got them yet. The V2 cameras. And they're going to show up tomorrow. So hopefully I get a chance to look at those tomorrow. Yeah, that's me. Okay. Thanks. Are we still on track to get all the parts we need to put together the other next group of prototypes. Yeah. I haven't received the speakers yet. So I'm going to check on that they were supposed to come in this week though. That's the like the biggest. I think that's the only thing really. Chris there. How's it go. Hey, do we have shown tell today. No, but we will on Friday. I did finish writing the API calls. So now the next step is to hook those up to the UI, but so I find I should have all that done. I did. There's two API endpoints now one is for getting a file, the information about the file to tag and one is to actually present the file to the UI, so it can be played. I also spent some time writing some SSH utilities and Salini using Paramiko because I found myself duplicating some code. When it came to doing SSH stuff between Salini and the Lawrence size server. So that's all coded. What else did I do. Yeah, I guess that was it. So anyway, I just need to write the code in the UI to to call those API calls and use the data to to display to display it on the screen. So the first thing I do is I get all the tags that are available, all the different types of tags, and then I randomly select one. And then once I randomly selected one, I go back to the database and find a file that has not been tagged with that tag type yet and has not been tagged in the last hour. I was concerned about, you know, with the SQL returning the same file over and over again if I didn't put something in there to, you know, to say, you know, don't, once it's been tagged, don't have it reappear again and again and again. So that's that's where the hour thing came from. So, so yeah, and that's it. That's basically it is pretty simple. The SQL is not complicated yet. But it might become that way if we get too too crazy. Any comments on that. Sounds like a good starting point. Okay. All right. So, yeah, that's me. Okay, so it sounds like you've got the pieces are in place and you're going to spend the next couple of days time together. Yep. So it looks like we can get this done by Friday. Yes, I will have. I'll have a show and tell on Friday afternoon and won't be. Won't be in production like I think we wanted to have it in this, this grant, but it'll be ready to start, you know, promoting to tasks. Okay. What's the difference between that and production? Just, you know, I wouldn't need to do the PR. I'll need to be reviewed. And then I'll want everybody to go out there and play with it in our test environment. Right. That sounds perfect. All right. About Ken, what's going on. Nothing good. So, I'm back and forth on the pull request reviews from Chris and guys. Thank you guys. But unfortunately, it's probably for naught. Since all that stuff's changing pretty quickly. So I don't think I'm going to continue on that. The wake word upload stuff is failing, but that's expected because there's no authentication on the URL I have right now. So when the real URL is out there, it'll pass. Just a note from your email, Josh, it looks like we're accumulating somewhere between 30 and 50 wake words a day. I didn't do any analysis on what, how many users that represented. But that was good to know so that we just can't break that URL. So I'm assuming the new URL will be different. I spent a lot of time looking at the USB stuff, Michael. I actually look at the Linux USB drivers, you know, we're open source, right? I'm still incredulous, but you're absolutely right. It's pulled and it's just weird and it has a really tight polling requirement. USB is odd in that all of the transfers are initiated by the host. So even inbound transfers are, hey, give me an inbound transfer. Which is odd, but anyway, it is what it is. So, yeah, whatever that means, long term for the switches, I'm sure we'll figure it out. I've already got some ideas on that. The last thing was I took a bunch of reading on some of the stuff Kevin sent me, verified this with Kevin. The stuff I was doing for sudo is not for null as the F control requires sudo as well. Why it would require sudo to go out of the USB port is beyond me, but it's very low level at C. Maybe it's going directly to the USB underlying interface rather than through some USB driver. But I verified with Kevin, you know, he actually pulled up the documentation that said that on Linux sudo is required. I tried to run it without it didn't. So I was looking at that and some of the stuff we have there. And then I actually started trying to, you know, get some stuff going on this and realize this SJ201 is not going to work like that. Even if I send out the F control stuff, it's not going to do anything because it's not hooked up like that right now. I'm assuming. So I'm looking forward to... Wait a second. There shouldn't be a difference between this and the next in terms of the XMOS chip. So are you saying that through the XMOS chip, we already have I2C connectivity to the amplifier in this? Okay. So I'm going to have to follow up with Kevin. I assume once he gets boards back, that's where he'll be at and he'll be able to give me some guidance on what exact commands will accomplish what. So that's where I'm at right now and I will probably get back on the VK tests tomorrow from the Celine side or the actual tagging process. Since I'm not expecting Kevin to have boards and have answers for me that quickly. Gotcha. So the hope that by going with USB interrupts, we would free ourselves from those low level driver problems. It doesn't pan out because we still need some kind of weird low level driver for the USB through the XMOS chip to get to hits I2C. Yeah. So I'm not sure how that's all going to look on the inbound side, but I'm pretty sure we're going to end up with a thread that's polling periodically for the switch status and generating pseudo interrupts or calling callbacks from there. Right now that's fine because the switches and everything are pretty much glued together, right? Like a switch is going to increase or decrease volume, things like that. There's one minor concern there regarding the, well not even with the signal. So the issue is if ever we want to allow users to give callbacks, there'll be some linkage issues there. Okay. Josh, how goes the update emails? I'm still working on Belina and getting it working. It's a management system for Docker. And I've got everything packed up. Other than realizing that I'm taking four to five hours of meetings every day, which I'm going to start putting a kibosh on. I have made some progress. I should have an update by the end of the week. If nothing else on how Belina works and then how it works relative to the evaluation criteria, which I did write down. So that's kind of where I am today. Hopefully I'll make some more progress between now and Friday. I'm in a working environment that's quiet. All right. Great. Okay. I don't really have much of an update. The next forever, the Mark twos is sorry, the SJ two of ones is still expected Saturday. Last I heard. And yeah, that's that's all all the news that we have on that front right now. So anything else that people would like to bring up? What's going on with the community? Yeah, well, a couple of things. I was on the new model and you said that it works better on the mark one and not the mark two. That's somewhat surprising to me. How does it work on your laptop? I didn't try that yesterday. I think it works. Works better. I'm surprised by that input. And then just the last thing somebody posted something about Python ID ease. Right. Yeah. Yeah, real menus. That's my input. I need it. I need something to help me debug. And now I see why you and okay, get along so well. Be good or we're switching to Erlang. I didn't think that has to. All of you, all of you hush. Tell us what you think in the comments down below. Which functional programming language would you like us to switch to? I did put some information out there about Python. Okay, yeah, thanks. Again, so I was just, I was curious like, how, what's going on with the community right now? Is there any updates you can give us on things that might be tough topics? Well, I mean, my attention has been very much on the on the lingua franca stuff. But what else has been happening? There's the porcupine wakewood spotter seems to have had an update that, you know, it's broken something at some point. So there's people working on that. Yeah. Is that still work in progress? Yeah, at the moment. So we've got a pretty good. Has anyone given that a go? No. I didn't want to be able to give a go to. Yeah, I'll throw a link in the chat. Yeah. There's a lot of instructions in JIRA, I think on the ticket. This year is so amazing. But yeah, so it's, it's sounding pretty good. I think we want to go back and get some new recordings for specific sounds. But I also want to make sure that so the sounds that it's having trouble with like soft sounds at the end of phrases. So I think it's a good idea to the way that the recording studio tries to, um, tries to, you know, package up its recordings. So I've, I've heard, I haven't looked into this, but I've heard that it sort of tries to detect where the end of the clip is and then and chop it for you, which seems like a really bad idea in some ways or like that we should do that. So it provides some grace period so that we don't actually chop the end of phrases off. It chops off some of the end of words and that's been causing problems. So I think, I think it's caused more problems with like soft sounds. So I'm assuming that it has a better, it can, it can more easily detect the end of a word when it's a nice hard like to, you know, whatever. When it's softer, like, then it chops a bit more of it off than it intended. So I think that is my complete hypothesis, you know, no, no data on that that's actually the cause of it, but we do want to go back, I think, and get some extra phrases recorded with some of those soft sounds, particularly at the end of the phrases. Yeah, but, but I'm using their voice locally. So it's, it's working. I can give you some samples if you want. Spin up my property very quickly. I don't understand. There you go. Was that a Russian accent micro? It's, I'll send you the link. It sounds, it's not as crisp as like I, you know, I think we could get, but it's sounding a lot better than it used to. That's, that's the key. And is the data, the data is the problem that we trimmed it when we, when we recorded it? I don't, I don't know. Like, I think that that was what David used to say. That was what David was saying when he was trying to retrain it. El Ticino, Ryan, Ryan's less sure about that. You know, because he's been able to get it a lot, you know, to sound a lot better than, than what we used to. But we're also using Takatron to now we're using Missila's TTS rather than the minute to TTS for this run. And so maybe that's part of it as well. But yeah, I'm, I'm still thinking that the data has a part to play in that. How does it compare to the Kassol voice? Do you think? American? I mean, yeah, like somewhat comparable. I think, I think it's, I haven't really done like a, you know, trying to jump back and forward kind of a thing. I've just been using it and it, it's, it's 100% usable. I'm kind of preferring it. So maybe that's, I don't know if that's just, you know, I, people expect female voices from voice systems and, you know, that's, that's a lot of work that a lot of people have done. And that's why all the assistants, you know, kind of do that female voice is the default. So I don't know if it's my, my, you know, weird, you know, being that just prefers a female voice or if it's actually better, you know. Yeah. All that work is being done by, by the community though, right? Yeah. Yeah. Okay. Bye, Ryan. Awesome. Yes. All right. Okay, well, thanks. We'll call it there and we'll get together on Friday and we'll have a cool show and tell Chris. And, oh, and I just received confirmation that the sj201s will ship on Friday. So. Awesome. All right. Thanks everybody. We'll see you in a couple of days. Okay.