Downloading file from R2 but unsure what the best method is.

[remote ginkgo]

Hey there, newbie here. Ive read several cloudflare docs on how to give users temporary access to a file using signed urls, however i want that the url is only accessible once and not for a short period.
The other method would be by using these object.httpEtag attributes but i really dont know how i can use etags in my application, all i could find about etags is that browsers store these tags to either use the file from the cache or download it again, but i dont need any cache at all, i always want to download the file no matter what (i understood the concept of etags, but dont need it, i want to download the file always).
My last idea would be to just get the file data within my worker and then just pass the data back in the response, but this kills the cpu time and isnt quite good as a free user.

Any idea or examples of a simple eimplementation to fix my issue? I dont wanna rely on any ai prompts but didnt find many resources online (probaby missed some good resources). A one time use link would be perfect.

[quasi flower]

My last idea would be to just get the file data within my worker and then just pass the data back in the response, but this kills the cpu time and isnt quite good as a free user.
As long as you return the body/readablestream, the worker is done and it doesn't consume any CPU time for the user to download the file
https://blog.cloudflare.com/workers-optimization-reduces-your-bill
(That wouldn't be CPU time eitherway, which is what Workers are limited/billed on)

But yea I'd just throw it behind a Worker. You may need Workers Paid if you are getting more then 100k requests/day but otherwise pretty cheap.
S3 Presigned Links just won't work, nor would R2 Custom Domains with HMAC or anything. You need something to statefully store the file has been downloaded, and you could use a Worker to do so, by calling an internal api, or using KV, or whatever, Workers can conditionally do it.

[remote ginkgo]

thanks for the info, however i think the way im currently doing it will still bill me or is this way right?

[quasi flower]

It still wouldn't be CPU Time used while you wait for the file to be buffered in, but yea that's a silly way of doing it eitherway and would explode if it was too large. All you need to do is return fileBytes.body in the response constructor

then it'll be streamed right through

[remote ginkgo]

alright thank you

also is there any limit for the maximum file?

[quasi flower]

R2 has limits like 5 TiB per object

[remote ginkgo]

i meant in the worker, sorry

[quasi flower]

streaming through a Worker has no exact limits though. Workers only have 128 MB of memory, but since you are streaming it through, it can be whatever size. I've served multi-gb files via Workers before from R2

[remote ginkgo]

ah thank you very much

[quasi flower]

Uploading to a Worker is a different story though

[remote ginkgo]

is there a api for R2 to do it directly without using a worker? e.g using account credentials

[quasi flower]

You'd have your zone/website upload limit (100 MB on Free and Pro, 200 MB on Business and 500 MB on Enterprise by default)

[quasi flower]

R2 supports the Same S3 API AWS S3 uses via S3 Compat

You can use presigned links to allow people to upload directly and bypass the upload limit through the Worker

https://developers.cloudflare.com/r2/api/s3/presigned-urls/

[remote ginkgo]

alright thank you very much