Unleash The Graph's data - Simon Emanuel Schmid | The Graph

does this work nice it works uh do Broan H thanks for uh being up already that early uh I like in Belgrade like um you know you guys know how to party uh that resonates with my lifestyle um but yesterday I I was kind of serious because I had my talk today so but yeah let's go into it um yeah my name is Simon or Simon or um I thinking ser you would say zimo no yeah okay um so I don't care about the pronunciation actually um it's just just uh yeah Simon zimo and today we talk about uh the the graph or like unleash the graph sta it's a little bit the marketing title the more nerdy one is like we talk about the idea of subgraphs as read oriented rollups does this work a little bit okay cool let's get into it so um who know knows what is the graph I do yeah so most most people nice okay so um the graph like simply explained nice it works this time with the if the animation is uh yeah it's just like takes all the data from the different blockchains and also from offchain data and then brings it organizes it in in an easy way so it's easy easily accessible or we can also look about um the architecture of adapt so we have the front end on top the idea here is like that's a decentralized front end so it's maybe distributed with ipfs or RV and then we have the subcraft that sits in between the front end and um and the contracts which are are on the lower L level and the contract after the blockchain data so that's roughly uh the system architecture that we we want to say so that that's actually is then the data layer of web 3 kind of the the big the big claim here um and it's true I mean it's like a decentralized Data Network it it runs in three years already or even a little bit more like as of now I think it I had to send in this presentation a month ago so like these numbers change but like uh there are 107 indexers 100 and uh uh uh 60 64,000 delegators that number went up a lot since um we trans we transferred the the protocol contracts from mainnet to arbitrum so it's much cheaper to be delegators so much more people become delegators everybody can be a delegator so look into it uh there are also uh 732 cators and the sub GRS that number goes up now very fast because like on the 12th of June the hosted service which was the centralized offering to demonstrate product Market fit of subgraphs will cease to exist so if you have any subgraphs and you're still using them or use any subgraphs in your application or your data science stuff then please upgrade you can talk to me I help you upgrade as of now uh 3.4 3.57 million query fees were generated or like sent through the graph Network um it's roughly $1 million us with the current prices so that actually gives a truly decentralized data so like the graph network is a maybe if you want to drop a boss word here like the what is it dpin decentralized P public infrastructure Network that actually works um in all Builds on top of subgraphs and we want to kind of like go into this in like what are subgraphs how do they work and why do we can we think about them as read oriented rollups okay so what is this subgraph simply said um let me try this again on top we have uh the database so subgraph defines a database schema then we have um a mapping that says how to fill that database schema according to the to to the data that we see on chain so so think about just a database and uh instruction how to fill that data base with onchain data so that's basically it so what happens then is like that on the blockchain we have all these transactions from all these different protocols um lined up and and we we share the block space right right I mean at least on L1 and and on the big l2s we will see where it goes with all these St chains but this is kind of a little bit the reality that we we share block space all the different protocols and but what we need is kind of like a a dedicated view like that we just look at one protocol at a time so this is what sucrafts enable uh they're fully um programmable we go into that later so the subcraft data flow as I said before we have um blockchain data first comes in then it runs uh mappings in assembly script or rust um depends if you are in sub crop or substreams land then it it rites to the store a database we can also read from the store and in the end we can send out queries to the customer that's roughly how it works um the cool thing about assembly script and rust um there is a lot of magic in it right uh these are T incomplete languages um I don't necessarily need to explain touring completeness here but I think you can think of it you can do whatever you want like they're fully yeah whatever like so for example one of the side projects that I worked on with some artists was to have a synthesizer that runs in the mappings of a subgraph uh and then according to I don't know like the block cash of the nft that get minted gets audio generated uh so you can do crazy stuff like this but also um this idea that came out already three years ago also uh kind of this idea of like say hey we can basically just build use the graph because like there is this um this determinism each subgraph is clearly defined um B basically as a mini blockchain so we just put very small stuff on chain and then like all the computation happens on chain so that lead has um that has created this so-called data Edge smart contract which basically is just some similar to rollups right they also just uh uh store some some State on on L1 uh we could have that data Edge where you can actually store whatever you want uh in in into the blockchain and then we can use something like the graph to index it and then even make sense out of it so what happens is that um logic moves from the blockchain or from Smart contract into into into assembly script if that makes sense so this um also like that this thing like moving logic or computation outside of the blockchain into another system that's um also now has like a lot of new um ideas around and then goes into this AI co-processing so there is another Co core def at the graph semiotic who who looks into this Co co-processing a little bit more um I explained the idea with with sub GRS as they are something that's already used but um if you're interested in this thinking um check out the gra.com blog there's there's a lot of talks about now how that's works with the Cod processing and so on okay so we talked about onchain data and and um tting complete mappings so that we can move logic to outside off the chain so to say but we also want to maybe have some offchain data um when you think about offchain data um it's everything that exists that's not onchain maybe but what is interesting and maybe also like to frame it a little bit what would be something that I want to publish or someone wants to publish could be uh airline free seats for Airlines they maybe want to publish this somewhere I mean currently I know there is some um I think Amad or some some some centralized service that uh kind of is in that market but we could theoretically think about just publish all of it on a yeah in a web threeway right uh we can think of threee Hotel seats pretty similar but um also an interesting example would be um with uh weather stations so we have a decentralized network of weather stations and we can put that data onchain and and then run computations and then actually um an example that we do at that to note is um server data so like access data to a server we can also then publish publish onchain so how would it work um so the IDE idea roughly is we go in the same uh steps um but now the mappings not only have access to the store and the blockchain data the mappings also have access to ipfs or RV data um and that's actually very simple so someone just uploads that data let's say on ipfs takes the ipfs hash and then puts it back on chain and then we have a subgraph that indexes that contract get that one and then it can start to download from IPS so that works actually pretty well um and with that as I said we came up with this um or like the data science team at came up with this uh so-called quality of service Oracle so when people send queries to the graph Network they go through gateways and from the gateways then to indexers um but so the gateways actually know which index is fast to respond to which query to which to which query actually and that's something that's important that we want to want to know in the ecosystem like which indexes are actually performing well um how that works is like the logs are then piped out of the gateways into red panda and materializing I mean like that's that's some data side stuff that I don't really understand but where I understand it back is like when it goes on chain so then they publish it to ipfs and then they use um open sapling um triggers to put it on nosis chain and on NOS chain it get index as a subgraph and so now we have a quality of service Oracle so whenever you kind of want to know which are the highest performing indexes which are the ones that have the quickest quer response times and so on um there is a there's a subg graph for it you can create on the graph Network um yeah the other thing that I said is my more like a hypothetical thing it's like if you would think about um weather stations so that could be like a DP network of independent weather stations everybody runs that at home maybe get some tokens to run it blah blah blah but they all upload their data um into on onto IPS or R gets anchored back on chain and would could feed maybe then that we go later into uh an oracle that would be like uh I don't know you create an open air here in Belgrade and then you want to know uh on chain if it rained or not because you want to trick Trier the the decentralized insurance um and that would might might be a thing because like if it's if it's a lot of independent people that run these weather stations then maybe we can trust the data more than it was just one uh agency that would give that weather information that could also be corrupted I don't know kind of a theoretical example here um but yeah reusing the pre the the properties of the graph Network that already exists and as I said already served that many queries um yeah as I said like um the that's about monetization so when we have data that we want to publish usually it's a problem that's why often times public data doesn't get published because then someone needs to have that API endpoint and that needs to be rate limited and maybe we need to have authentication maybe we want to have some some billing on top just to pay our server fees and so on so forth but like with what I just said is um it's very simple it's almost for free to publish that data and make it accessible and then basically sell it through the graph um Network yes so back on chain as I said like these both things we have like these oracles for maybe weather and quality of service um we need to bring it back on chain and that's actually something that that already works there are the chelato web functions but also the chain link um chain link functions that can get data from the graph Network and put it back on chain um so so there are different use cases so for one would be like the one that just goes with onchain data that would do like complex comp computations across on chain data that's gets in the subcraft and comes out with just one value that maybe I don't know let's say for um for a def risk management thing that that could be piped outside that one or as I said with offchain data that we even um get in weather data and then we have an onchain uh contract that can say did it rain or not according to to that setup so that would be the the full the full um cycle that's why that idea about rollups comes right it kind of like with with rups we also um want to offload computation stage from L1 to make it cheaper and faster but you also want to bring it back on chain like for for the or like back on an L1 for what what what concerns L1 so to say um so that's the the roll up thing and when you look at think of subgraphs as I said then the the data gets up gets protest and we can bring it like very um we can bring it down on chain if we need um questions are later but um yeah thanks for the contribution um yeah so then there's of course there's this question of verifiability um so the data that gets indexed currently there is this proof of indexing going on um it is uh index when they see data then they computate um a hash on top of that data put it at back on chain to say like they actually I index that data they can they start to U talk to each other and um although they they compete they also talk to each other did you see the same data on yes no should be um should be very deterministic you can think of but interestingly it's not like different configurations of different uh ethereum node or L2 nodes then could be that sometimes an event gets missed and interesting stuff like this happens and we start to uncover it by having that system um then also with the query when we send a query to the graph Network each indexer responds with a query attestation and says like I with all my Stak that I put into the network I guarantee that it's true right and then if that if that could proven wrong it can get slashed there are fishermen so there's this kind of gem Theory going on um but obviously we want to move forward into more like strong um guarantees which would be verifying indexing querying um and there is First Steps going on like simple data extraction from from chain um with when we when you get that mer Merc Roots up into that indexing layer then we can have start to have some um some some uh some checks unfortunately Jason RPC doesn't really help with that so there's this reinvention of data extraction going on um that also other teams do but like what in the graph ecosystem it's don't the fire hose um so we can have maybe this these guarantees later on and then like all these verifications then go layer on layer we trying to work up but um that's future and I think that's it all already um you can scan my QR code follow me on Twitter my DMs are also open if you have any questions but I'm also still around now for questions and you can also try to find me uh yeah as long as I'm in Belgrade thank you so much so we have some time for a couple of questions so if there's any questions in the audience feel free to raise your hand don't be shy don't be shy we're totally cool dudes we're there are no wrong questions I I I think I think the baby asked all the questions like it was a good question anyway like like thank you thank you so much um Simon for for being here for sharing your knowledge give it up for