[core] How does a job invoker provide secrets to the module runner at runtime"

[AquiGorka]

The idea here is to have an open discussion on how to give modules some inputs that are considered sensitive.

[richardbremner]

I think we need to look at the big picture here.

The request comes from "how would we run lilypad build on the lilypad network". I still have some concerns with that, like running a build to fix the network when it's down.

But perhaps the real question is, "how does a module invoker (do we have a better term for that?) provide secrets to the module runner at runtime". It seems like a ws callback (or https) could be a practical solution. We could define this url in the module spec so that there's a standard way for module authors and module invokers to achieve that.

[developersteve]

This needs to start at the client side when the job is requested and signed using the module requesting the user wallet or an on-chain encryption service.

The actors in the process, the job initiator, will need to easily pass unencrypted inputs, which are then encrypted and sent to the network to be processed. The RP selected to run the job then decrypts the input. Once the job is run, the output is encrypted and sent back to the requestor. The validator can then decrypt the results to validate the job.

As a side note, this process would need to be light and fast so that it doesn't take up too much processing time and is easy enough for a requesting user to pass inputs.

[richardbremner]

@developersteve that was my initial thought too, but the job initiator doesn't know who's going to run the job so maybe I'm missing how they'd decrypt it?

[AquiGorka]

@developersteve that was my initial thought too, but the job initiator doesn't know who's going to run the job so maybe I'm missing how they'd decrypt it?

Ahhh true, and if we wait until the deal is agreed to get the public address and send the encrypted secrets, at that point in time we might have to do it via a ws callback.

running a build to fix the network when it's down

I'd definitely go with an approach that has a plan b: keep the github actions that can deploy around but these can only be triggered manually, if the Lilypad network cannot run the CI+CD pipeline, we can always manually trigger the github action as a fallback.

How does a module invoker (do we have a better term for that?) provide secrets to the module runner at runtime"

Updated the title, thanks for the suggestion

[developersteve]

https://infisical.com might be an option to investigate