Cartesi Compute allows you to implement a function in which you receive a set of inputs and it provides an output. So you can think of a Compute DApp as a computational oracle that tells you the result of a computation.
How it works
Cartesi Compute SDK allows Cartesi DApps to specify and request verifiable computations to Cartesi Machines. Additionally, the SDK provides tools to facilitate and reduce the cost of inputting data into Cartesi Machines.
What follows is a summary of the architecture and the components that are involved in running a Cartesi Compute Application, meaning some DApp that makes use of our SDK. It is convenient to start with the description of the typical ingredients involved in a decentralized application. Namely, the blockchain node and the client software.
Cartesi is a second layer solution.
At this point, an overview has been given of what constitutes a Cartesi computation, who are the parties involved, and what the software components of Cartesi Compute are. It is important to understand better what events happen during the execution of a Cartesi Machine.
Although an extensive documentation of Cartesi Machines can be found here, one may choose to skip this reading and jump right away to their usage inside the blockchain through Cartesi Compute. For that, it is enough to regard a Cartesi Machine as a black box that executes computations.
Having discussed the concept of Cartesi Machines off-chain, capable of booting a Linux operating system and loading heavy-weight libraries, one naturally wonders how this will ever be stored or executed on the limited environment of a blockchain. The simple answer is that it won’t be.
Having informally discussed how Cartesi Compute represents Cartesi Machines on-chain, one can now describe in more details the API for requesting and retrieving computations in Cartesi Compute.
This section describes the main ingredient of the on-chain Cartesi Compute infrastructure.
This section describes in detail the Drive _inputDrives parameter of the instantiate call.
After going through the last section, the reader is already able to specify drives if the data was available to the caller at the time of instantiation.
A relevant limitation of the Cartesi Machines as they have been described until now is the size of their input drives.
When users interact with their blockchain DApps, they are free to manage and run their own blockchain nodes if so they wish. In a common scenario of the usage of the Ethereum network, DApps are accessed via browser and blockchain transaction requests are carried out by Metamask. The user signs the transaction which is typically sent to a remotely hosted node, such as Infura.
Broadly speaking, the Cartesi layer-2 platform architecture should be perceived as blockchain-agnostic, given that in principle any network could use Cartesi Machines to move complex computations off-chain without compromising on decentralization.