Summer Fit Check, Cron Schedulers, and Sample Apps

CJ: What is up, Colin? How's it going?

Colin: Pretty good. How's how's the week?

CJ: It's great. This is the final week of school. The kids are done. Officially done with elementary school, which is just mind blowing. So, yeah. It's all intermediate school and beyond from here.

Colin: Very cool.

CJ: Lots of end of year activities. They've had like field days and field trips and plays and all kinds of stuff wrapping up the year. And,

Colin: it's summertime.

CJ: I think they're, yeah, it's summertime. It is full, yeah. Full force summertime. Last night we did a little fire pit in the backyard and they were jumping on the trampoline and just, yeah, it feels awesome. So,

Colin: cool. How about you? Getting out, doing anything? Fitness wise?

CJ: we're yeah. Lots of. Lots of fun stuff. So we went for the first time ever to visit New York city a couple weekends ago. And Nicole and I booked classes where you can take classes at the Peloton studios in NYC PSNY. So we went to went to the studio and did two live rides with Camila Ramon and one was a reggaeton ride. And one was like a low impact ride in the studio. It was so awesome. Like it is. A full on experience to go there. It's like an Apple store type thing. And you go in and they have all these different bike studios that when you're inside them feel like super, super legit, almost like movie studios. And there's, I don't know, it was, it was cool to see. The instructors, we bumped into a handful of like Peloton instructors and the, when you see them on camera, you think they're just normal. Like they've got like the same energy that you have, you know, you're like what's yeah. What's mind blowing is that in person you can tell that their energy is like level 9, 000 and they're like just so pumped and so on. It's. It was really, really impressive to watch. She was like speaking in English, speaking in Spanish, like singing lyrics, giving instructions, dancing, like every single like millisecond of the entire production. She was just like on point and it was, it was, Amazing. It was so cool to see. And yeah, it was, it was tons of fun. It also like inspired me to put more energy into like my own YouTube videos. Like they're not, they're not exercise videos obviously, but like definitely It was inspiring to just be like way over the top and super energetic and super, you know, enthusiastic about whatever's going on because I think it just makes for better

Colin: Yeah. Are those rides that are, are those being cast to people who are at home?

CJ: yes. So their lives, if you're at home, you could take it or you can take it later to you. Like they they record them and take them on

Colin: that's kind of like what we've talked about. Like when you record anything, you kind of have to dial it up so that it comes across at a hundred when you're watching it, because when you record it, if you record it at 90, it's going to be like 70. And so you got it to go at 130. And that's something I try to practice here. The other thing that reminds me is like the Aeros tour and Taylor Swift. I think both Taylor Swift and then Beyonce for her things, like they will sing their songs on a treadmill and just like training, cross training, like doing, you know, taking it to 9, 000 and so that they can, you know, be physically fit to pull off these amazing shows and have, you know, such like. physicality and presence in, in all the things that they do, which takes a lot and it's hard to see when you see it like just on a screen. Yeah.

CJ: don't really appreciate it nearly as much until you're in person. And yeah, you got glitter flying all over the room and it's like,

Colin: Peloton is an interesting one to me because like I, so as like, as far as like stocks and stuff, I mostly do indexes, but Peloton is one that I had been buying just because it was so low for so long, but obviously it had a huge surge. During COVID right. And then a lot of the cycling companies, the tech companies around fitness and running and all these things have kind of taken a huge hit since, since things opened back up, but like this experience you're talking about is like hard to pull off for somebody else to do. And I see this in the hotels now where you can just like, they have Pelotons where you can sign in with your own account and like you can train, you know, you're not. Using some foreign equipment, it's the brand and people want to use it. Cause I think Peloton users that I've talked to, like, they love it. I've actually never used one. We've talked about you and your treadmill, the Peloton tread. But it's interesting that it's still like so low and still taking those hits. I'll be curious to see, like, if they go more. this like experience route or like we're moving into these new apartments and the gym just has like a row of Pelotons that you sign in with your account, which might be the thing that gets me to sign up for just the like the pass so that I have an account to log into when I'm in the gym. But we'll see, like, there's a lot there that makes that interesting to me in terms of like, now I have an account I can take with me when I'm traveling and still like get the gym in, in a really easy way, instead of like being in a city that I might not know where to run or where to go do stuff.

CJ: Yeah. There is, I think having a Peloton at the hotel is a pretty big draw. In fact, they have hotel finder. one Peloton. com, which we can drop in the show notes. And it's like this tool that helps you find hotels that have Pelotons. And. That was like my first experience with it was, going to San Francisco for three weeks to onboard to Stripe and the place that we stayed had a Peloton. I was like, what the heck is this thing? Let's give it a shot. You know, tried it out. And then you know, several months later I was like, I think. that would be really good for us. And so we got one right before the pandemic and ended up, I mean, I've done over 1100 rides now. So for us, it's just been like a really great tool for exercise.

Colin: Yeah. There's the gamification element and tracking and stats for people who want stats. Yeah.

CJ: And yeah, there's lots of different programs, lots of different, there's, I feel like they, they have something for everyone, right? There's stuff for the, the metrics nerds, there's stuff for the people who just want to be entertained or have fun, the stuff for beginners, stuff for people who want to be super hardcore about it. And Yeah, we've, we've just had tons of fun and it was awesome to go to like the Mecca, which is like, you know, the, the the studio. So that was, that was awesome.

Colin: Very cool.

CJ: I know last time that we were on, we talked about RTO, the Reno Tahoe open that you were going to run. And I saw some pictures on Instagram. It seems like that was tons of fun.

Colin: it was a good time. It was my 10th RTO. So I finally like added up all the legs that I've run and I have like four more slots that I have to finish. So it'll take, if I do it the normal way, it'll take another four years to have completed all of the, the legs. I think there's like less than 50 people who have done all of them so far. So we'll see. I think there's, I could do another, like I've done a few ultras where I've done multiple likes in a year. So like if I do two ultras, I could be done in two years, but I think we'll take it nice and slow. The, the doing it for 24 hours, it took us 25 hours this year. You only really get like one hour of sleep. And speaking of metrics, like you can try to get sleep whenever you can. It's just like people you're hyped up. You just ran. You are trying to eat when you think you should eat. And then you're like, when should I have coffee? When should I, it's just like your body is like, what are you doing to me? And you know, I've been using the whoop band for almost five years now. And every, every year you can see when I do RTO, cause it's just like the worst, like they have a joke about the 1 percent club and that's when you have a 1 percent recovery. And so I got the 1 percent badge again this year for just the worst sleep. And you're also running. So I had like a my run, the hardest run I had was eight miles. Six of it was uphill at one o'clock in the heat and then two miles down after that. So got it done. Felt pretty trained. And yeah, I kind of, this will go into what we talk about later, but, you know, we've, we always like to talk about work life balance, but I'm still trying to find more ways to get away from the screen and, and get out there and, you know, obviously I can just run or use Peloton or things like that, but these, these kinds of like tent pole races are like things that keep me. To it, because if I'm not trained, I'm going to suffer more. So having a few events planned out has really helped me do that. So always looking for like the, what the next couple are. And of course the memes and the jokes on, on Instagram and stuff are like, once you hit 30, once you hit 40, it's like. It's time to run a marathon, right? You're, this little thing in the back of your brain, you're like, I should start trail running. These little things that happen and I, I think those are probably good midlife crisises to have. So so yeah, if anyone's listening and wants to get into running or trail running or pelotoning, you know, shoot us, shoot us any questions you have, but really just, Get out there and start like, you know, trail running is like hiking with maybe a little bit of jogging on the downhill. So it doesn't have to be like what you see, you know, all the pros doing like, make it, make it your own.

CJ: So, one thing that I was wondering about with RTO is, I thought that they, you had like the team follow you in the van, right? But there's sections where you're not running on the road?

Colin: Yeah.

CJ: so then you like run away and then you meet up with them at a

Colin: So my leg, the eight mile one, you can't have someone follow you. Cause it's like a dirt road up a mountain, it's like a mountain fire road. So there's a few like that. There's a downhill from Virginia city down Geiger grade. It used to go down Geiger grade, which is a road, but it's like it's super sketchy, scary road. So now you actually go down a fire road, which is like. Super like just rocks and scree and like a lot of bloody knees at the end of it for a lot of people who fall. But that one, the van can't follow you obviously down that. And then the van can't follow you on mine for the one that I did. You have to carry water because it's usually during the middle of the day for most people. And it's, it was thankfully not as hot as it is today, but it was hot. It was like 85, which. It's not as hot as it could be in other parts of the world, obviously, but like right now it's 95 and I would not want to be doing that in 95. So yeah, so for the most part you have two vans and the sleep that you can usually get is when the other van is running, you try to get your sleep in, but you still got to also go to where the van's going to meet you. So you have to spend some of that time driving. to the next spot and then, you know, either crash on the side of the van, like just not crash the van, but take, take some sleep, like in a sleeping bag next to the van at the, the trading, at the, the switch point, or like what we do is we rent a motel six and our van. uses it and then the next van, van two comes and joins and takes that room while we go run and then then they meet us at the next spot. So it's nice to just have a little spot to do a little reset and move on.

CJ: Nice. Gosh. So the, every year that you've done it, have you done different segments or is there like a segment that you run every time or you're trying to collect them

Colin: Yeah. So there's 12 total and I've done it 10 times, but I've, I've done it I've done the ultra three times. So like I've technically run one, two, three, like 16 legs. So I just, I've done a lot of the same legs multiple times. So I subconsciously have favorites, I guess, cause I didn't realize that. So I went through and looked at all the ones that I'm missing. And now the next couple of years, I'm going to try to knock those ones off. So I have two legs in van one and two legs in van two to do. And we'll see how that goes.

CJ: Nice. And when you say ultra, is that a separate race from RTO or does that just mean you're

Colin: Yeah, the ultra is so normally it's done with 12 people. There's 36 legs, so the ultra is usually 8, 6, 4 people. There was a year where one person did the entire thing. So that one was pretty gnarly, they did it, they started earlier. And it took them like two days, I think, to do it. So, yeah, so I've always done it with six people. So you end up doing double the legs, which is a little bit more brutal cause you don't have the other van. So you don't have any sleep really. Like you can try, but it depends on how you cut up, which, which legs each person does. And a lot of spreadsheets, a lot of like timing, a lot of like stopwatches and stuff, but it's good fun. And, and it also like, you know. Gets me again, just to like, I am not going to be at the computer that week. It's like, and then you got recovery and all of that afterwards, but good times.

CJ: yeah, yeah, that's great. Great way to get out, especially in the Reno Tahoe area. Get out and see some.

Colin: It was beautiful. So that's the fitness update. What are you what are you building? What are you working on?

CJ: Oh man. We're building a lot at craftwork. We, I think we've talked about calendars a bunch of different times, and we're finally to a place where we can interact with a calendar for building shifts and timesheets and tracking approvals and understanding who's where, and just this week we started adding. The concept of different markets, meaning like different cities that we're going to be able to like expand to. So if you're looking at a calendar for Charlotte, you'll see like, here's all the projects that are happening in Charlotte, North Carolina, and all of the crews and painters and drivers who are in Charlotte and who's booked, who's not booked, what's happening at the project. Are we painting? Are we pressure washing? Like, is this an interior, exterior, et cetera? So that's been pretty fun getting that over the line. And this is another example of a tool where we'll sort of collapse down a third party tool that the team has been using so they can just use our our software. So it's been Almost actually it was just, just this week. It's been one year since we broke ground on the rails app and we kind of did like a little Spotify wrapped type situation about you know, like where are we at with the code and how many PRS we do and how many lines of code have we written and things like that. And it was pretty crazy to see. Like we basically replaced a whole chat system. We replaced an entire, like Google calendaring system. We replaced a whole bunch of features that were being used in notion. We replaced two way email. We replaced like yeah, like lots and lots of these different little. Things that and this one is like one of the, one of the last ones is a tool that we use called deputy for managing shifts and timesheets and clock in and clock out and stuff. So slowly but surely we're building all that functionality into the product. And

Colin: a few, a few episodes we talked about API platform risk and this sounds like a lot of those things you're just checking off and reducing the risk. And I know we've, we've talked about build versus buy a lot, but you know, part of that equation is so much of the like integration points and the headaches that you also adopt. And like you guys are, you know, able to own that capacity, which is good. And of course, as I hear you list off all those things, I hear like the whole internet, but Rails, Rails doesn't scale. CJ Rails doesn't work. How has Rails been?

CJ: it's great. I love it. I love working with it. We've definitely hit a few like scaling things. Like most of the time at the, well, yeah, most of the time, if something is slow, it's because we have some naive query that's happening somewhere and it's not being cached or we like just don't You can see that pagination or, you know, there's just like little, little things here and there. Like we show a badge in one of the navigation menus that tells you how many unread messages you have. And we just like re query every time you load the page for. Like show me all the messages you have read and then give me the Delta between that and like how many other messages there are. And so yeah, like low hanging fruit stuff where we need to add like a little Redis caching layer. But other than that, it's been super fun. We've been using hotwire and turbo a ton and it has worked surprisingly well. So. We just added a, there was a request for a feature. So we have a command palette thing using ninja keys. I don't know if you've seen that, but yeah, senior like similar to in linear where you can do like command K or you can use the slash key to pop open a little command palette. We have a command palette for jumping around and there was a request to like add a feature where in the command palette you could see recently viewed projects. And I was like, at the same time I was thinking it would be nice to have the experience that you have in Google docs, where you have like, you know, everyone's face who's viewing the doc right now in the top, right. And it was pretty fast to get up and running with action cable and just showing who's there real time and then collecting up all those project views into some models and then spitting that out as like you know, recently viewed projects. So we have kind of like a LRU cache type situation for. The command palette. So you can quickly bring up recently viewed. Pages and you can also see who's on the page that you're viewing and yeah, I've been really pleased with that. I think we have followed convention, it's really benefited us. So yeah, I wrote a blog post that we can leave a link to with just like some reflections on the last year and also just like things like rules of thumb that we've sort of picked up for rails over the years that folks might be interested in. I think in particular, I've gotten very opinionated about enums and like not to use them certain ways. So if you're curious, like I don't reject that up,

Colin: That's, that's a good one. I'll have to read it because we have the similar things. We've, I've come, there's like the rails, there's a lot of different opinions about that in rails. And then I've now learned a lot about this in terms of we use a lot of like integer. Enums. And the nice thing is you don't end up having to rename them later. Like when you're like, we're going to reuse this or change it. You don't end up with like install v2 final final. Right. So you'd be like, well, now zero is this instead, or one is this, but yeah, it can be enums or there's a lot of foot guns there.

CJ: Totally. Yes. Yeah, I think there was, there was, there was a video that I made on YouTube and someone in the comments was like, why don't you just like, it's a best practice to always have like store integer values for enums. And so I just like trusted this random person who left a comment and I was like, Oh, that's good feedback. I'm going to use that now that someone on the internet said it was a best practice. And I like, didn't really like look into it, but I was just like, Yeah. I internalized it. I was like, okay, well maybe this person must've been right. And like, surely for performance, it's better to store them as integer values instead of the string values in the database. But the reality is like, as soon as we had downstream services hooked up directly to Postgres for our like business intelligence tools, now like everyone needs access to your Ruby model, like your actual like Ruby code to know like

Colin: Yeah. What is one, what is

CJ: is. Yeah, exactly. And so now we have these like giant select case statements in all these third parties and anytime an enum value changes, now you have to go update it in the third party too. So

Colin: like you have to have a response model or something in an API to make sure that you translate the enums to a nice string value. Or you, I guess you could return the integer, but now you're telling developers into like either create a library that then translates those into whatever those values are. Yeah.

CJ: there was, there was other, like the other story that I might tell is that like we started out with what I thought was going to be a small set of enum values. And over time it was, it was to represent types of services that we provide basically. And it was like, Oh, we're going to paint the walls. You can paint the ceiling, paint trim. And over time the team wanted to like adjust and shift with those meant and add more. And so as we started adding more union values, we got to like, Over 25 or 26 or something. And then we started to like encode things into the string value. That was like, Oh, paint underscore square foot, underscore brick, underscore sighting. And that meant that it was a paint thing that we're going to do. And it was square foot based. And it was this other thing. It's like, Oh, actually like the enum key in this case is encoding like a bunch of data that should be probably in its own model. And would be much easier to work with. So yeah, lots, lots and lots of learnings there.

Colin: Very cool.

CJ: Yeah. What else

Colin: got some sidekick stuff going on.

CJ: Yeah. So in the past, I don't know if you remember, there's like a Heroku add on called Heroku scheduler. Yeah, I used to use this thing all the time. This was like my go to way to like set up any like periodic background job. Meaning, yeah, it's, A lot of times we'll build background jobs that just do some processing in the background based on a trigger, meaning a request came in through a controller and then you kick off a background job, or a webhook came in and you kick off a background job. But there are things that you want to do like every day at four o'clock, or you want to do it, you know every Friday at six or something, or maybe you want to do something every minute so you can go through and check your database to see if you need to do any, you want to do. Sort of data sanitization or whatever. So all of these periodic task type things, I used to solve them by writing rake tasks that would kick off background jobs. And then Heroku scheduler was this plugin by Heroku that lets you set a cron style. Thing that would kick off your rake task on a schedule and there was like a little ui that you could use but now that we're on render we didn't have access to that plugin And so I started looking into these a couple of different alternatives. So whenever the whenever gem was one that was part of jumpstart already, but I liked sidekick cron for a couple reasons one it because we're already using sidekick You And sidekick web, it adds like another tab to sidekick web where you can click on the cron tab and you can see here's all the jobs and when they're scheduled, when they ran last. And you can even like, Enable and disable them from the UI and you can kick them off from the UI. So I really liked the way that that fit in. And so set that up on my personal side, just to like experiment a little bit with fetching these podcast episodes so that it updates on on the website. And then also I set it up in buckets, our little finance app that we've got going on to just start fetching some data that's not coming in via webhook for a couple different things. So property price values from some property APIs and yeah, I think it's, it's, it's been serving us really well. So we started using it internally to send push notifications to our mobile app. At the end of every day to like remind people, Hey it doesn't look like we've seen a progress report from project X and it looks like you clocked in there today. Like, can you send a, send an update? So,

Colin: Nice. Yeah, we made, we made heavy use of that at the subscription underwear company. Right. We had a lot of Heroku scheduler going on for subscriptions or like it's time to run everyone's subscription. Cause the way Shopify worked, we had Tell Shopify when to run the subscriptions. It was kind of maddening. Even though I think it was using Stripe under the hood, which is another story for another day, but and then at orbit, we use sidekick cron with sidekick, like. We were constantly looking at that screen and just seeing like, literally like the one that was always backed up was like integrating with discord because the right limits, it was always like, okay, how many sync jobs are running? But it was super nice to be able to say, run this every hour, run this every day at four, run this whenever. And we would usually schedule them. I guess that's probably where they, they came up with the name for the gym. But we would often schedule around. Like we, we had built a. An HTTP client that also was aware of the queuing and the cron so that it would stay within the rate limits and like respect it. And then like delay job if we were going to hit the limits or, you know, whatever was going on. So I find myself missing some of those tools in JavaScript land. I know you could just write your own scripts in JavaScript, but like the idea of like, these are rake tasks that then create jobs. That then have workers that run in those jobs. And it's like in JavaScript, you have to kind of just go do all of that. I just found a bowl MQ that I'm looking at for some stuff. There's a few blog posts that I found where like top five message queues and. Background processors to use. Cause I'm working on kind of like a best practice for billing in a bot or an app on discord. So like kind of what are all the events you need to listen to and don't do it in band and store it in a job and then process it and stuff like that. So I was like, okay, I normally reach for. Sidekick. And I was like, okay, we're in JavaScript land. I'm sure there's probably a way to do sidekick with JavaScript, but it's probably not designed for that.

CJ: Yeah, I have no idea. I have not seen that. I haven't seen Sidekick used with JavaScript, but when we were when we were first building with Next. js, we used this tool called Ingest with, it's like n with two n's. And at the time, I, I thought it was pretty, it was a pretty new company. And this is like their, their entire company is built around like being a service that provides background job handling there were a couple of things I really liked from ingest. One was the logging and like just keeping track of when your jobs ran. And then they also had some like. Interesting concepts around, I think it was around like trees of jobs that could have interdependencies and things like that, which can get kind of messy. But yeah, it worked great for us for a while and we used it to send emails and handle a few things before we set up rails. So but I, yeah, I remember looking at bull And a couple others that were just like, how do you host this? Where do you run it? Like, what is the worker? Like if I, if I deploy my next JS app on Vercel, like is my bowl MQ thing happening on Vercel or somewhere else? And then it was like, do you just set up a serverless function that runs your background job as a separate thing? And then like in your main job, you're like to schedule your background job. You just hit the serverless function. And then, yeah, I don't know. There was, I remember it being one of the many frustrating things in JavaScript land being that like every single piece of the puzzle is like, where do I get the third party SAS product that solves this problem? So,

Colin: Well, and it's tricky when you start looking at the renders and flies and stuff. It's like, if you're running in this like quick to deploy thing, like it's like, it's like why Heroku had an add on store and why they had Heroku scheduler. And I'm sure Chrome, like they have probably some features, but when you just hit a key, like a worker, you loo there's so many things like, do we have observability? What happens if it fails? What happens if it needs to call some other ones? And there's like a fan out and there's all these different things that you have to think about. So like, yeah, looking at Ingest's website, it looks like they've maybe started leaning towards AI jobs as a service, but still, still a worker system, but very heavily on that. LLM chain and like, usually you're doing lots of requests. It's like chatbots and things like that. But, you know, it looks like flow control, logging, observability, recovery tools. You're doing all that yourself if you don't use a tool or like, I think some of it, what's nice about Sidekiq is there's so many plugins to Sidekiq. There's like retries and there's back like exponential back off and you can define all of that stuff. We used to use a company that is no longer around called iron worker or iron or iron. io. That was very similar to, this is like over a decade ago, but very much like, I do not want to think about where these jobs are, where they work and you can deploy go and Ruby and all these different ones to them and let them do that, but that's like a whole, it's a whole world. Mm hmm.

CJ: Yeah. Yeah. I think in Django we use DJ celery or like a J or some, some sort of celery thing. And I remember George had to go like super deep on debugging some lock issues where like we were, we were writing to. Some cue and celery had to pull stuff off the queue. It's like not simple to get right. And so, yeah, I do appreciate that sidekick is backed by a actual business. And it seems to have lots of great support with different plugins and features and oftentimes the stuff too, that comes with the paid. Plans for sidekick other than support, obviously, but a lot of the features that you can get from Paid are also available as community packages or a lot of it is just ruby So you can kind of treat it as a ruby object and go into it and write your own. I think we built A debounce like our own like debouncing thing. So if you kick off several different jobs that all have the same arguments to perform and they're all item potent, then it'll debounce them until you stop creating jobs for like a minute and then it'll run the last one. So yeah. Great little tool.

Colin: Very cool. Yeah, on my side, I've mostly, like, been working on those bots, like I was talking about. Mostly just carving out some best practices, figuring out what our docs are missing, figuring out what things, like, kind of cursive knowledge that we have. Like, what are we missing? What are we doing? Discord is interesting that we don't have our own SDKs, so we, like, I'm writing everything using just, you know, Pure API calls right now, but we'll probably also do it in a few different libraries just to, so that people have some examples, the libraries have pretty good support and guides and stuff like that, but there's some things where like the libraries don't always implement all the things that we offer, so just trying to show off, like, what is our interpretation? Maybe what we recommend is not even necessarily how it's written in the docs. And so when you do. Get that back in line. When things just change and need some updating.

CJ: So when you're working on a sample bot in JavaScript, are you're using like a third party, like community JavaScript library, or are you writing like just fetch calls or how's

Colin: Yeah. Like I'm usually using like Axios or something just to have a nicer little rests and point or fetch if I want to just have something without any dependencies. Mostly just because then the reference implementation is without any specific library so that library devs can see it, developers can see it, and then that way if you're building something and you're doing it in another language that's not JavaScript, you can just see what the control flow is on the JavaScript side. Are we going to end up with like examples in a bunch of other languages? Maybe, I think like most of our bots are probably JavaScript and Python. So those are like the two big languages. Discord, RB is really cool as like a Ruby one. We use that at at orbit. And yeah, there's, I mean, there's 10, there's dozens of of SDKs. So like there's someone who has one in R and in Crystal and in Ruby and Elixir. And so it's,

CJ: do you get telemetry on like which ones are most popular or is it kind of like anecdotal? Like,

Colin: Yeah, they're supposed to send like a user agent. So we have a good sense of like what the top libraries are. And a lot of the really large bots have their own clients and, or forks of popular clients just because when you get to large bots, Really large bots. They have to deal with the gateway and sharding and like limits on the gateway of how many like if you are running multiple instances of your bot connecting to pools of guilds and servers, like there's a lot of overhead where a lot of our examples are for bots that we just assume you're not going to get to that issue. And like our newer bots mostly use HTTP and there you're getting like webhooks instead of having to do an open web socket to the gateway. So there's a little bit of a movement that way. We don't support webhooks for everything that the gateway has. It's only for very specific, like the slash commands in discord are powered by webhooks usually. Or you can do them via the gateway too. So some of this is like, okay. Now that we have this, like, should we be offering more events? Should we be moving to an HTTP future? And, and cause the nice thing about that is you don't have to do sharding and you could have one instance of your bot running that. Takes those and puts them into a job queue, for example,

CJ: and it, are these examples all hosted on GitHub somewhere that we could check

Colin: They will be, yeah, there'll be in the docs. So I'm working on like a refresh of a lot of our docs content, just trying to get rid of some of the older stuff. We've got a lot of things that are like, this is deprecated. So it's like, okay, we might just say it's been deprecated for a while. We're going to shove it into get hub. Here's the get commit if you want to go read it again, but. We're going to get it out of the docs. So yeah, there'll be in the docs, discord. dev.

CJ: Very cool. And are there, when you're talking about like building apps around best practices, like how are you, like, how do you go about figuring out what patterns people should implement? Are you looking at like, ways that the bigger bots have done it or talking to like the engineers internally or, yeah, like kind of what's your approach to figuring out the, the right or the way that you want to recommend?

Colin: this is kind of like a side quest for me that I haven't really told anyone that I'm doing. So what I'm doing is my interpretation of what I would do if I was an external developer, because that's the failure point is like, what are we telling people to do, and then I'm going to take it to the team and say, okay, I implemented this. Please poke holes in it. Like we'll do like a demo day type thing. What would you do differently? What are the concerns that I'm not, what are we, what are we blind to that we aren't telling people? What am I, what did I also do that wasn't in the docs because I know that I have to do this thing. So that's a note for me to update the docs. I think for a lot of developers, like they want to see code and they don't want to just read because they're not going to read usually. We hope that they read, but the code can be very self documenting that way. And then we can also take PRs if the community thinks that something could be better. What I'm trying to figure out is like, The discussion we just had about workers and stuff is that's when you get into opinions. It's when you get into tech stacks, right? Like when I start thinking about like, here's where we should implement a caching strategy, we could just say, choose a strategy of your. Your own and just not show an example, or we could do what I'm trying to figure out is like the same app. Do I do it multiple times? One super naive in band requests. You can run it and it's going to run fine. No dependencies. The next one's like, okay, this is using bull MQ. And here's how you would save these two. Job queue, process them, and why you would do that. But then, of course, you get the why aren't you using my favorite queue system. And then the other one would be like, okay, here's another one using Redis as like a cache, and here's how we recommend, you know, doing your cache and validation and stuff like that. Those are not things that we document. Like, it's not our job necessarily to document how to do caching as a job. Concept, but I think if we expect you to have a good time that you should be caching or we should be offloading those things, then I think we should tell people that because otherwise you're going to hit rate limits, then everyone's going to go ask, like, why am I hitting these rate limits? And so there is a little bit of that, that I think we can do to make people more successful. And then we run into the, okay, now I want this in my language. Now I want this in. You know, some flavor of a library because a lot like most people are not going to do raw HTTP requests either. So like, what is the Discord. js example of this? But my hope is that if we offer these that then Discord. js can translate that into their own version. Or they can tell us like, Hey, we don't think that that's the best way to do this. Some of the SDKs, they're really, they're like more like thick clients than thin clients that they do a lot of the caching for you or like the discord RB one it has. A really cool way of keeping you within the rate limits. And so like, it's aware of how much you've used what's available. It will intentionally delay your requests and things to make you fit within that. So it's still possible to hit it. Like if you do something, you can slam into it really hard. But there's some of that stuff that's like not going to be captured in like a generated open API SDK, for example.

CJ: Mm hmm. Mm hmm. Yeah. One thing that came to mind while you were explaining that was the, there was like a period at, at the end of my time at Stripe where I wanted to give people my opinionated implementation of how I would do something with Rails. For a Stripe connect integration. And I wanted to show like, here's how you should handle your web hooks. And here's how you should like farm stuff off to the background. And here's how you should do X, Y, and Z. And it really, we had a hard time like finding anywhere that it might fit in the docs, in the proper docs. Cause in the docs, it's like, you just want to tell people the high level stuff that they care about from that's like tech stack, agnostic, whatever from Stripe And it was yeah, so it was like this, we got into this situation where it made more sense to write them as blog posts that were just posted on dev. to or my own, like my own website or whatever, because it's like, it's actually just my opinions, but about how I would do it. And with those like additional things like sprinkled in. So where in the docs you might say like, yeah, we recommend using a caching strategy here in like the blog posts. It's like, okay, here's how I did it with caching in this specific stack or whatever. But yeah, I think I, I did a bad job about. Like doing a lot of collaborations with other dev rel teams, but there were other people on, on our team that were good at saying like, Oh, Hey, we might want to do a caching strategy here. Let's do like a partner video with some other, like some company that's like a caching company. And so then it's like, okay, here's how you could use ingest in the context of a discord bot. That's like sending the thing. And then you kind of like work together to make a live stream or you work together to make a blog post or whatever. Yeah. I don't know. Kind of,

Colin: Yeah, that's a good

CJ: it's tough to get

Colin: a good call out because I haven't gotten that far yet But this will probably be like like a dev 2 article or it could be you know Let's go do it live on Twitch or a discord dev YouTube channel, which we do not have right now, but we might one day Those kinds of things so we'll think about that. But yeah, the the tricky thing about partnering Is we found this even with open source implementations of like a certain job queue is then like our docs are public. So if we implement one, we're going to get PRS for everyone's favorite, whether it's their proponent of that one or their DevRel for that one, or they want GitHub contributions to the discord repo. So, you know, we get all of them. We, cause we did that for activities. We, we picked a third party. Multiplayer backend. And the cool thing is that company has that since built their own discord example, that's way better than ours because like it's from them, they understand their tool better than we do. So we were probably going to swap that out for theirs. Cause a lot of people were like, why did you pick that one? And you know, why aren't using mine or this paid one? It's like, well, it was open source and our team clearly had used it before. So it was more of like, we want to give you an example instead of make you do it yourself. But, you know, some of those opinions, I think what I would like is that the docs should at least make you think about like that, like that you should think of a caching strategy. And then maybe that's a, Maybe it doesn't make sense to do a call out to like implementation there, but if it feels more human, like you can have the super technical docs with a fun little bubble. That's like, we recommend a caching strategy, go check out how CJ did this and rails or whatever.

CJ: Mm

Colin: you can kind of cross promote, which would be fun because I think like at the end of the day, like especially with discord. Like we should have some fun with it. It doesn't need to be this like cut and dry, like you know, manufacturing guidelines for implementing XYZ. It's like it's discord.

CJ: Yeah. Speaking of fun, I, the little touches that have been popping up in D script lately have been pretty cool. Like,

Colin: The name of the AI assistant was a strong choice. I don't know any context about why it's called Underlord. Is that what it is?

CJ: Yeah. Under the under Lord, you go to the under Lord to have it automatically do stuff for you.

Colin: Well, in a world where, like, almost every AI assistant seems to be named, like, a woman's name, which is, like, problematic all to itself. Right? And, like, or they're trying to make them human feeling, and now you've got, like, the Underlord, and

CJ: Yeah,

Colin: I wish I was in that, that marketing meeting. 'cause naming things is always hard, but also so fun

CJ: I know. I'm sure someone was just like, what if we called it like the overlord? And they're like, no, no, no, no, not overlord. That sounds scary. And they're like, under Lord, like sold. We'll call it

Colin: But then, you know, that whiteboard also had like, what if we called it Dan or DS Script? Dan

CJ: Yeah. Yeah. Or like the Descript dog. Oh, like, you know, like a lot of people want to have a, a mascot. Situation,

Colin: for sure. Well, we'll be using under Lord to edit this episode. So

CJ: yeah. Brought to you by the Underlord. Oh gosh. All right. Why don't we wrap it

Colin: right. Good chatting. Where can we find all the notes and all the things?

CJ: Yeah. Head over to build and learn. dev for links to all the resources to the stuff we talked about today and yeah, we'll catch you next

Colin: All right, bye friends.

CJ: Bye friends. All audio, artwork, episode descriptions and notes are property of CJ Avilla, Colin Loretz, for Build and Learn, and published with permission by Transistor, Inc. Broadcast by