And on the free plan it definitely is pretty limited in customization of how it's defined

would I even need one tho?

Yes. But you would be still have to run the worker. You see the requests will still come through. The worker would still be executed. Even if to check that your account balance has expired. So it would still cost you. And If someone is spamming the api. Like with 100req/s then your worker and all that code for doing the checking of the limit is still getting executed. So you’ll will be billed for all that requests.

Adding one wouldn't hurt

Just to prevent spam burst requests

i have mine set at like 100 per 10 seconds (or maybe 10 i forgot)

I see

that's fine, I don't need user-level rate limit customization.

$1.40/million requests seems really nice considering my work uses AWS where that's the minimum basically

Is your api highly dynamic ?

wdym?

Since you said in your api you’re just fetching the request from another website and presenting it in a nice format. So what I ask is that is that very dynamic ? Like maybe per request it would change drastically. Or like there could be a time limit for when the content updates. Beacuse if that’s the case you can use cache and reduce requests a lot.

not likely to change too often - I'm already using cache, but it's unlikely to reduce requests too much because it's not very often that a user will repeat the same request

Is the content per customer ? Or can the same content be served to multiple customers ?

multiple customers

Would the customers use the API directly from the frontend ? or from their own servers ? Or it can depend on the customer and basically unpredictable ?

from their own servers

So, it would never be frontend ?

never

yep

Okay,

Then @p0 I believe you would benefit a lot from external services in your worker such as PlanetScale. $1 per billion reads are quite cheap, and if you design your sql efficiently, you can make it 1 read per execution of your worker. So, then that way you don't have to use either KV or DO. And PlanetScale is atomic by nature, so you don't have to worry about it.

I'll look into that

thanks for the help

You're welcome.

Check it out here. https://planetscale.com/.

alr

but wouldn't it still have to write to increase the value?

Yes, but writes are comparatively priced. But once you have hit the limit, you are just doing reads thereafter. So no more writes. Something like this: x <- read if x > 0 then do work write (decrement) end if

writes are 1.5 per million