Swaroop Hegde I PowerLoom Announcement and Scaling Trust | Pragma Waterloo

foreign [Music] so for our last demo I want to bring on swarup to talk about power loom now this is a really interesting project because there's a lot of amazing Origins here which we'll dig into for how this got started and not only that this is something that matters to all of us whether we're developers or not on as we kind of use more things on chain so I won't ruin the surprise I'll have to talk about power loom so without further Ado please welcome suru thanks so much Gothic um it wasn't so much of a secret now that you're hyped it up I'm like a little more nervous now I thought open sourcing this like a week or like I think a month ago was probably going to be easy uh but no um anyway my name is uh sarup I'm one of the co-founders of Paulo um I live in Vancouver a very different side of this country but it's exciting to be here especially with its connection to both ethereum and 8th global I don't think I could have picked a better opportunity to showcase what I've been sort of working on with my small team for the last two two and a half years um to give you some sort of context um so I've been in the space for since 2017 building um developer tools for ethereum so we started with a project called we used to help onboard webto developers a very different time I think we had less than 100 000 devs and maybe this is probably what the hackathon venue would have looked like I think it it was as well so um and I think that that's what our goal was um we uh we built that for a few years and eventually went on to help the small uh Network called Matic which you now all know as polygon and um they grew quite a bit and this is actually a real graph of uh when we were running all of the infrastructure in I think some around 2021 um it was pretty crazy to sort of do something at that scale uh but one of the things that you know we noticed uh was that a lot of the uh dabs the D5 projects and pretty much everyone else uh using on-chain data would pull that data sort of run it through a lot of third-party services or like some sort of In-House databases and caches and present that to their end users which sort of Beats the whole purpose right uh why we are in this space um and I like how I think Karthik Alia asked about what is truly decentralization so you know if you sort of take shortcuts uh there's no purpose and we ourselves were one of the main people right like we were running 90 of that infrastructure and we challenged ourselves um what could we do to change this and I think it was one of this hack and Karthik I just tried to pull this from some really old blog post uh here's the zoom call and maybe a story from the other day um where I sort of presented um I don't think it would receive in power loom at that point we were just testing out one of the ideas um it was quite the demo I hope my today's demo is not like this uh and we ended up um testing this prototype that we had and um we ended up then started working on this protocol so um the premise is very simple uh the ID band Power Loom is to aggregate data from on-chain uh this really depends on the use case how you transform or sort of splice data around so it could be coming from one contract it could also be coming across Shades uh but the end result is a sort of a snapshot and that snapshot is something you don't have to query right that's the big difference between everything else that's out there and what we do is we use cryptographic proofs uh to maintain their data integrity and the biggest advantages because we sort of store this on well-known sort of battle tested Technologies like ipfs and filecoin uh anyone can sort of pull that data instantly in a really cheap way um and even if let's say in a network where there are certain parties that are sort of doing all this computation any period on that Network should be able to pull that and I'm hoping I can demo that today um so uh maybe we'll sort of jump into um how to sort of run a problem node uh I've done a really simple one-click setup uh been working on it all night so hopefully it runs today uh live I wish I had recorded the demo like the last speaker I think that would have been better but let's try this um and actually before I jump into that maybe give you guys an idea of um what use case are we working on so I'm hoping some of you have used unisop we sort of talked a lot about swapping tokens and stuff so uh we built this use case which is similar to start its own Within um so when you use unisol the behind the scenes there's a lot of data that sort of gets stored on chain right so which is what people use to make decisions and then um like in this case like tally is sort of helping people to sort of protect those swaps from happening for different doubts but it's this final data that unisop shows on the info pitch uh now this is pretty complex now even if you sort of try to load this page you'll notice the time it takes in spite of it actually using a third-party service to pull the status this is not coming from on chain the source is on chain but the funnel output is sitting on some uh and not to badmouth any particular uh competitor or the third party I think they're working towards sort of solving that so what we have done is we have built this sort of use case on top of our protocol to Showcase what we can do and we're going to try and uh launch this uh node so okay so let's head to our oh I already have the tab pulled up so let's head to the main repository this is our uh deploy everything is open source so you can actually try this uh locally on your Docker but I don't want to bring down the Wi-Fi here so I'm going to try and deploy this on what GitHub calls uh code spaces um the advantage of that is it doesn't matter what kind of computer you have right you should be able to deploy this remotely on that instance but for full decentralization I would highly recommend that you use Docker and do this locally anyway so um let's jump here I already have one code space running here so maybe let's try to deploy there it's got a nice name fictional guacamole okay now I wanna I'm actually feeling hungry thinking about it um okay cool that actually loaded pretty quick so let's clear this up um so this essentially uh runs a lot of different components uh which I think maybe there's not enough time today to go uh in detail uh but in terms of inputs it expects way a few things so you just need to provide an RPC URL which could be in the ideal state or run some sort of a light ethereum node uh in this case I'm just using this free provider to show you how it works and you would have to probably set up a key uh so any sort of evm wallet ideally something that's fresh for our setup so I've set up one here already and that's it I think that should be enough uh I think I've hidden oh I've hidden the private key downstairs because uh this is completely gasless you don't need any gas we are using um sort of a lead to to run these transactions using a layer so let's actually run the script so it's a one-click setup hopefully um yeah all the images are already pulled so it's going to run your ipfs node as well but you can use third parties like uh infuro and some others uh and while this is loading up let's actually go and check out what it's going to do behind the scenes so we have this test net that is just went live uh it's not fully open to everyone but you can definitely read uh for for you to participate um there's a small form that you just need to fill the only reason we haven't allowed everyone is you can easily bring on the network with a few hundred participants if you just Spam it because there's no gas involved at the moment but we'll open it up uh in the coming months for everyone so as of now just go to powerloom.io and fill in a form and come find me either today or during the hackathon and I'll give you access right away so these are all what we call as uh the fundamental uh projects so these are like sort of modular in the senses so in this case you can sort of see if it makes sense let's try to zoom in and understand what's going on so for this particular use case there's a few pairs that we have selected um this is what we are calling it like a light mode but you can choose pick and choose depending on data market and eventually once we have incentives then there'll be uh completely dependent on demand and Supply so this is like a pair contract of a certain pairs so this is similar to what you see here these are these pairs that run on Unison upright so we're tracking bits of each data and then combining that into a final snapshot so one what you see here is for each Epoch so what we do is each Epoch is essentially linked to this uh let me try to pull this I actually had a tab open there we go yeah so it's like a spanner block how it's right now so it can be completely dependent on uh whoever is running the service but as of now it's every 10 ethereum blocks for the sake of just uh for from a human perspective it's easier to understand what's really going on so once an Epub is generated all the notes in the network receive that through an ebm smart contract consensus that we built and hopefully we'll have something better uh in our next test net and when you're uh node receives that it does this all of this computation and submits snapshot against it and when they reach consensus we see this sort of a finalized state right so in this case we had four peers uh reaching consensus and so let's see how many come in so I'm hoping by now our node is up and running let's see yep everything is good to go so let's see and let's look for this node here and if it has started submitting so so far it has not submitted anything yet but if I refresh it I'm hoping I think the Nets it's the docker is not able to connect it but anyway I'll try to show the final output of what it looks like um so this is what you see when you run your own node so if you remember I had this info page pulled up here so this is what is running from currently from a hosted service that unisop uses but in our case every node has a copy of this and the way it does is it Aggregates all of those small bits of projects and turns it into this final output and the good thing is all of this is sitting on top of ibfs so if you actually try to open this through either your ipfs node or through cloudflare again I think it's always with the the network might be a bit slow and pulling in but you can go to the website and you can check this on your phones as well so you'll be able to verify that exact same snapshot um the CID matches for every node in that uh Network so when we take all of these projects we turn it into this one use case the idea in the future is developers should be able to build whatever they want so this is just a early Showcase of what we've been uh we've been working on and let's give it one last try and see if that there we go perfect okay it has finally submitted one and hopefully that should get uh finalized in before the next debock hits so let's see how many there we go do we have three uh snapshotters I think it waits for at least 51 percent or it waits for the uh time window so at least that worked that's good yeah that's it that's how you can launch your problem note so uh you don't need any expensive computers or anything else and I think that was our goal for the last two years and I'm excited to show this and working uh big key Point awesome thank you so much and please do check it out amazing live demos are uh are hard especially when you're launching a test net um so uh it went better than the last time and uh maybe it's a good time for you to tell the story here of what that Zoom screenshot actually was and how your backend was perfectly appropriate for I think they were the most painful five or ten minutes of my life I remember it was I was also stuck in India during covet at that time so um we were in extreme lockdown in the sense we couldn't go out of powerhouses so whatever internet connection I had probably some middle of the night as well um and it's funny because I think it was Zoom that throws it and I remember when I bought this computer um that's the first thing I just zoom work um I uh oh yeah it was actually Apple pay I was trying to do this uh entire demo um again live as usual I think I just enjoy doing live demos and um but we managed to sort of get it to work and that got a lot of attention and that's what I tell people that it doesn't matter sometimes even something going bad can happen no there's no such thing as bad press I think but it's super interesting because we kind of realized through in your case through previous hackathons what some points of failures are and and why these things matter and and given how much we're talking about let's move everything to on chain data Integrity matters data availability matters in even though your focus was here on the test net and how easy it is to kind of get started on snapshotting uh can you tell to everybody here in the audience what you think the main benefit is of somebody running a power loom node or using this over any alternate data fetching methods yeah even um like we are calling it testnet purely for the consensus aspect of it the data that I was actually showing even that final demo that we see it's actually live uh that data can be verified um the only thing is because there's no incentive that there's that's why we call it things as testnet um I I think there was a gentleman earlier today uh when Jacob and I were there he asked um what is adapt and I realized um because he was trying to say what are you going to present and I thought nothing matters I need to go back to the basics right so um these are all like nice words um but in terms of when you explain it to a person who may not come from our space um the the key uh element is that even the sort of transform data right so we have this on-chain data which is expensive we talked about it I think you covered it uh with the in the earlier conversations with cryptokitties and flow uh we cannot be storing this file data again on it right and at the same time if we store it on ipfs who watches the Watchers right like that's been my it's it's one of the my favorite Turtles all the way down um and I think that's what we've been trying to solve um uh as of today I'm not saying it's perfect uh but what we have managed to ship so far and what I managed to still demo uh is ensures that whatever that use case is so in this case it was just unisol data but it doesn't matter right it could come from Sushi swap or it could even come from Cross chain so the final data is one the hash doesn't change because thanks to Matt actually oh that's all there is right there's just a hash of that uh Json or text file and what ensures the Integrity is when you have more and more in the network so that's what we focused on making sure it's so cheap that anyone can run this uh system that just solidifies um the Integrity of that data so when should somebody be using this over like the graph or something else which is more like a real-time querying of that same on-chain information right I think uh we do have a slightly different approach to like and I don't want to just pick on the graph uh you know they've been my customers and I think they were the reason how I sort of got thinking like hey this needs to be solved in a better way um theirs is more query driven right you're still um indexing the data and I if I'm not wrong at least it used to be uh stored on a postgres uh sort of a database um but with us there is no such thing there's the file data is already computed so even you can serve that data tomorrow without having to sort of query some sort of a centralized database because the problem is even if you index you know consensus you still are trusting one peer to return the result of that query in the right way right so that's something that's not a challenge with us but that's also um a bit of a drawback because it's expensive for us to pre-compute everything and sometimes latency trade-offs uh latency in pre-compute only uh post process ours is instantly available that's the good thing because it's actually very quick so you can refresh the page and get that in it's when you sort of try to do any complex querying and you haven't thought of that use case before then the nodes have to go and try to calculate that and read some sort of at least one uh Capital consensus so I guess one of my questions here is like we kind of saw this thing we understand how this is different than a querying language but what kind of new applications does this enable or what exactly is now possible that that wasn't easy or possible before I think at least when I first sort of got into the space and even before we got building all those rpcs we didn't even have a storage a reliable storage Network there was ipfs but it was more like a hobby project and I think that's why that connection to hackaths and uh you know the file coin team is so special because I believe they went mainnet just a month or two after that uh Zoom call of ours so that ensures that we have so here's the thing right we are also not Reinventing any like we're not trying to run our own chain we're not trying to we're actually using ebm stuff and all of these existing Technologies so the good thing is we are leveraging what's already out there so we don't need a new set of people to run new infrastructure file coin miners can run this ethereum proof of stake like I myself Runner node as well so I could run this as well uh the the only difference would be is because it depends on if there are more and more use cases there would be a more demand for someone to run dedicated infrastructure to run that uh transforming of the data in between and I think that's that's made it easy for us because we don't have to reinvent anything for our use case awesome well congratulations on the launch today and it looks like people who are interested in Computing data or pre-computing data and putting it on chain for the hackathon we're moving forward come find swarup in the audience today or anytime this weekend and uh use some free data key Value Store as a service yeah absolutely thank you