A DO droplet will absolutely suffice until you reach a large volume of traffic. Even then, migrating your data to a larger host is no problem.

Reply to this note

Please Login to reply.

Discussion

Ty

No problem! If you use something that utilizes a SQLite database, it will give you a ton of flexibility to scale by being really portable and easy to migrate into something more robust like postgresql

If you can filter spam, then it’s likely pretty good to scale as your grow.

Otherwise spam is around 10X non-spam events, if they add your relay or you sync from large relays.

Are there tools to filter spam?

I think Nostr-rs-relay and some others have a gRPC hook that you can run some check before publishing is allowed. Paid member, or content or rate limit, etc.

I don’t know of anyone else using it, however I wrote a couple ML models and have a decent dataset for training. I use it successfully in production - however it only works for events with enough content, not reactions, reposts, meta, etc. Bayes model works with fewer resources than the FastAI model.

https://github.com/blakejakopovic/nostr-spam-detection

#[5]​ 可δ»₯ηœ‹ηœ‹θΏ™δΈͺοΌŒη”¨aiζ₯θ―†εˆ«spam

ε₯½ηš„