Help using rust-fil-proofs

[moogamouth]

Hello, I want to use rust-fil-proofs for my project. I want to generate and verify proofs of storage, but I do not understand how the library works. I would greatly appreciate help implementing it. Also, could I please join the Slack chat? Thanks.

[sueun-dev]

I wouldn't start directly from Lotus for this. The current rust-fil-proofs README says the Filecoin-specific proofs crate is wrapped by rust-filecoin-proofs-api, and that API crate is where the public seal, post, and update modules live.

So for a project that needs generate + verify flows, I'd start there and use small/test sector sizes first before touching production sizes. The parameter/cache setup matters too: the rust-fil-proofs README calls out /var/tmp/filecoin-proof-parameters or FIL_PROOFS_PARAMETER_CACHE for proof parameters.

I only checked current public source/docs, not a local proof run. Can't help with the Slack invite, but for implementation help I'd ask with the exact proof type you need: sealing/PoRep vs WinningPoSt/WindowPoSt.

[moogamouth]

@sueun-dev Thanks so much for responding. I know that Lotus isn't the place for this, it's just the only place with active discussions. I don't really know what proof type I need, that's part of the problem. I just want to have a server and a client, and the client can send a publicly available file to the server, and verify at any time that the server is still locally storing a unique copy of the file. If the server tries to download the file from elsewhere when challenged, or delete/edit parts of the file, it will fail. The client should not need a copy of the file to verify this. The thing I'm building is decentralized, but for now, as a prototype, I'm basically building a toy centralized version of filecoin.

[moogamouth]

@sueun-dev Could you help me please?

[rvagg]

You could probably learn from the lotus, lotus-miner and venus codebases which all use filecoin-ffi to link in the proofs. Also the Curio codebase does as well. ChainSafe/forest is a Rust implementation of a validator node but not a miner, but it might have some learnings too. Your best bet is (a) learn from what existing codebases do and (b) use AI to help digest it.

[moogamouth]

This would be excellent advice if the codebases weren't massive, sprawling, and confusing, and if AIs weren't dumb. I tried doing what you said and it was a corn maze that AI could not help me out of. This message sounds a bit snarky, that's genuinely unintentional. Also, it seems like Curio and forest may lack a lot of the functionality I'm looking for.